JP7686965B2 - Inspection system and inspection method - Google Patents

Description

This disclosure relates to an inspection system and an inspection method.

In fields such as FA (Factory Automation), an object is illuminated while being imaged, and the appearance of the object is inspected using the obtained image. Conventionally, a method has been known in which an image is captured multiple times while changing the lighting conditions in order to improve inspection performance.

For example, in the photometric stereo method, the normal of the object's surface is estimated using multiple images obtained by capturing images multiple times while changing the direction of the light source. This allows the surface irregularities of the object to be inspected without being affected by dirt on the surface.

Furthermore, Japanese Patent Application Laid-Open No. 2016-105044 (Patent Document 1) discloses an image processing device that captures an image of the return light from an object each time the object is sequentially illuminated with P types of illumination light.

JP 2016-105044 A

When the object is moving, the position of the object in each of the multiple images obtained by capturing the image multiple times differs from one another in the conventional method described above. Therefore, the conventional method described above cannot be applied to a moving object.

The present disclosure has been made in consideration of the above problems, and its purpose is to provide an inspection system and inspection method that are capable of inspecting a moving object using multiple images with different lighting conditions.

According to one example of the present disclosure, an inspection system includes a plurality of illumination units that illuminate an object, a polarization camera in which unit areas including a plurality of polarizers are repeatedly arranged, and an inspection device. The plurality of illumination units irradiate illumination light with mutually different polarization states. The plurality of polarizers transmit light with mutually different polarization directions. The polarization camera captures images with the plurality of illumination units simultaneously turned on, thereby outputting a plurality of polarization images respectively corresponding to the plurality of polarizers. The inspection device inspects the object using the plurality of polarization images.

According to the above disclosure, each of the multiple polarized images corresponds to an image captured under lighting conditions that are primarily those of the lighting unit that irradiates polarized light that forms the smallest angle with the polarization direction of the corresponding polarizer. In other words, multiple polarized images corresponding to multiple lighting conditions are acquired by one-shot imaging. This makes it possible to inspect a moving object using multiple polarized images with different lighting conditions.

In the above disclosure, the multiple illumination units are arranged such that the azimuth angles around the optical axis of the polarization camera are different from each other. The multiple illumination units include first to Nth illumination units. The multiple polarizers include first to Nth polarizers, where N is an integer of 2 or more. The polarization directions of the illumination light of the first to Nth illumination units are parallel to the polarization directions of the light passing through the first to Nth polarizers, respectively. The inspection device generates a normal image indicating the normal direction of the surface of the object from the multiple polarized images, and inspects the object based on the normal image.

According to the above disclosure, the polarized images corresponding to each of the first to Nth polarizers are captured so that the light from the illumination unit that irradiates illumination light parallel to the polarization direction of the polarizer is the strongest. In other words, the illumination unit that has the greatest effect on the polarized image among the first to Nth illumination units is different from one another in the multiple polarized images. The first to Nth illumination units are arranged so that the azimuth angles around the optical axis of the polarization camera are different from one another. Therefore, multiple polarized images with different conditions regarding the azimuth angle of the illumination light are obtained in one capture. Then, normal images are generated from the multiple polarized images. By using the normal images, the unevenness of the surface of the object can be inspected with high accuracy.

In the above disclosure, N is 4. The first and third illumination units are arranged symmetrically with respect to the optical axis of the polarization camera. The second and fourth illumination units are arranged symmetrically with respect to the optical axis of the polarization camera. The difference between the first azimuth angle at which the first illumination unit is arranged and the second azimuth angle at which the second illumination unit is arranged around the optical axis of the polarization camera is 90°. The multiple polarization images include first to fourth polarization images corresponding to the first to fourth polarizers, respectively. The inspection device generates a first normal image indicating the magnitude of a component along the direction of the first azimuth angle in the normal vector of the surface of the object based on the first to fourth polarization images. The inspection device generates a second normal image indicating the magnitude of a component along the direction of the second azimuth angle in the normal vector of the surface of the object based on the first to fourth polarization images. The inspection device generates a shape image indicating the shape of the surface of the object based on the first normal image and the second normal image, and inspects the object based on the shape image.

According to the above disclosure, a first normal image showing the magnitude of the component along the direction of the first azimuth angle and a second normal image showing the magnitude of the component along the direction of the second azimuth angle are generated. As a result, even if the direction of a scratch on the surface of an object is random, a change in shape corresponding to the scratch appears in the shape image. Therefore, scratches on the surface of an object can be inspected with high accuracy.

In the above disclosure, the angles between the polarization direction of the illumination light of the second illumination unit, the third illumination unit, and the fourth illumination unit and the polarization direction of the illumination light of the first illumination unit are 45°, 90°, and 135°, respectively.

According to the above disclosure, the first to fourth polarized images correspond to images captured under lighting conditions mainly consisting of the lighting conditions of the first to fourth lighting units, respectively. Furthermore, the influence of the illumination light of the third lighting unit on the first polarized image can be minimized. Similarly, the influence of the illumination light of the fourth, first and second lighting units on the second, third and fourth polarized images can be minimized, respectively.

In the above disclosure, the multiple lighting units are arranged so that their elevation angles relative to the object are different from one another.

According to the above disclosure, multiple polarized images with different elevation angles relative to the object as lighting conditions can be obtained in a single capture.

In the above disclosure, the multiple illumination units include a first illumination unit that irradiates illumination light along the optical axis of the polarization camera, and a ring-shaped second illumination unit centered on the optical axis of the polarization camera. The multiple polarizers include a first polarizer and a second polarizer. The polarization directions of the illumination light irradiated from the first illumination unit and the second illumination unit match the polarization directions of the light passing through the first polarizer and the second polarizer, respectively.

According to the above disclosure, the first polarized image corresponding to the first polarizer indicates the brightness of light that is irradiated from the first illumination unit and specularly reflected from the surface of the object. In other words, the first polarized image corresponds to an image obtained under bright-field illumination conditions. In the image obtained under bright-field illumination conditions, the brightness of the area where the scratch exists is reduced. Therefore, by using the first polarized image, scratches can be inspected with high accuracy.

The second polarized image corresponding to the second polarizer shows the brightness of the light irradiated from the second illumination unit and diffusely reflected on the surface of the object. That is, the second polarized image corresponds to an image obtained under dark-field illumination conditions. In an image obtained under dark-field illumination conditions, the brightness of the area where dirt is present differs from the brightness of its surroundings. Therefore, by using the second polarized image, dirt can be inspected with high accuracy.

In this way, a first polarized image suitable for inspecting for scratches and a second polarized image suitable for inspecting for dirt are obtained in a single imaging session.

In the above disclosure, the multiple illumination units include multiple ring-shaped illumination units that are concentric around the optical axis of the polarization camera. The inspection device generates a phase image that indicates the angle between the normal direction of the object's surface and the optical axis direction of the polarization camera based on the multiple polarization images. The object is inspected based on the phase image.

According to the above disclosure, since the polarization states of the illumination light from the multiple ring-shaped illumination units are different from one another, the amounts of light that are irradiated from the multiple ring-shaped illumination units, specularly reflected from the surface of the object, and transmitted through each polarizer are different from one another. Therefore, stripe patterns with different phases are captured in the multiple polarization images. In other words, multiple polarization images with stripe patterns with different phases are obtained with a single capture. Then, the phase image generated from the multiple polarization images indicates the normal direction of the surface of the object. Therefore, by using the phase image, the unevenness of the surface of the object can be inspected with high accuracy.

In the above disclosure, the plurality of illumination units further includes an illumination unit that irradiates illumination light along the optical axis of the polarization camera.

The above disclosure makes it possible to reduce defects in the stripe pattern in the central parts of the concentric circles, enabling more accurate inspection.

In the above disclosure, the multiple illumination units include a first illumination unit that irradiates illumination light along the optical axis of the polarized camera, and a second illumination unit that irradiates illumination light from the back side of the object. The multiple polarizers include a first polarizer and a second polarizer. The polarization directions of the illumination light irradiated from the first illumination unit and the second illumination unit match the polarization directions of the light transmitted through the first polarizer and the second polarizer, respectively.

According to the above disclosure, the first polarized image corresponding to the first polarizer shows the brightness of the light that is irradiated from the first illumination unit and reflected specularly from the surface of the object. When the light is incident on an area where a scratch exists, it is diffusely reflected. As a result, the brightness of the area where the scratch exists in the first polarized image decreases. Therefore, by using the first polarized image, scratches can be inspected with high accuracy.

The second polarized image corresponding to the second polarizer shows the brightness of the light irradiated from the second illumination unit and incident on the polarized camera. Therefore, if the object has light-blocking properties, the shape of the object's outer periphery can be clearly recognized in the second polarized image. Therefore, by using the second polarized image, burrs or chips on the outer periphery of the object can be inspected with high accuracy.

In this way, a first polarized image suitable for inspecting for scratches and a second polarized image suitable for inspecting for burrs or chips on the outer periphery of the object are obtained in a single imaging session.

In the above disclosure, the multiple illumination units include a first illumination unit that irradiates linearly polarized illumination light and a second illumination unit that irradiates non-polarized light. The multiple polarizers include first to third polarizers, and the first polarizer transmits light with the same polarization direction as the polarization direction of the illumination light of the first illumination unit. The second polarizer transmits light with a polarization direction that forms an angle of 45° with the polarization direction of the light transmitted through the first polarizer. The third polarizer transmits light with a polarization direction that forms an angle of 90° with the polarization direction of the light transmitted through the first polarizer. The multiple polarized images include first to third polarized images corresponding to the first to third polarizers, respectively. The inspection device generates a composite image by high dynamic range synthesis of the first to third polarized images, and inspects the object based on the composite image.

According to the above disclosure, the non-polarized light irradiated from the second illumination unit is uniformly transmitted through the first to third polarizers. On the other hand, the linearly polarized light irradiated from the first illumination unit is transmitted through the first polarizer but is not transmitted through the third polarizer. In addition, the intensity of the light irradiated from the first illumination unit and transmitted through the second polarizer is 1/2 the intensity of the light irradiated from the first illumination unit and transmitted through the first polarizer. Therefore, the first to third polarized images correspond to a plurality of images captured under illumination conditions with different illumination intensities. In other words, the first to third polarized images with different illumination intensities are obtained in one capture. Then, the first to third polarized images are synthesized in a high dynamic range to generate a synthetic image. By using the synthetic image, the inspection accuracy of the object is improved.

According to one example of the present disclosure, the inspection method uses a plurality of illumination units that illuminate an object, and a polarization camera in which unit areas including a plurality of polarizers are repeatedly arranged. The plurality of illumination units irradiate illumination light with mutually different polarization states. The plurality of polarizers transmit light with mutually different polarization directions. The inspection method includes a step of acquiring a plurality of polarization images corresponding to the plurality of polarizers by imaging the object using the polarization camera while the plurality of illumination units are simultaneously turned on, and a step of inspecting the object using the plurality of polarization images.

The above disclosure also makes it possible to inspect a moving object using multiple images taken under different lighting conditions.

According to this disclosure, a moving object can be inspected using multiple images with different lighting conditions.

1 is a schematic diagram showing an overall configuration of an inspection system according to an embodiment of the present invention; FIG. 2 is a diagram showing an example of the arrangement of polarizers in a polarization camera. FIG. 2 is a schematic diagram showing a hardware configuration of an inspection device. FIG. 1 is a diagram showing an example of a conventional lighting device used in a photometric stereo method. FIG. 13 is a diagram showing the arrangement of a plurality of illumination units according to a first inspection example. 2 is a diagram showing an example of the polarization directions of linear polarizing filters 12a to 12d. FIG. 11 is a flowchart showing an example of the flow of an inspection process in a first inspection example. 1 is a diagram showing the relationship between the illumination units 10a to 10d and a normal vector n at a point P on the surface of an object W. FIG. 13 is a diagram showing the relationship between the traveling direction of light emitted from an illumination unit 10c and the normal direction of a point P when the traveling direction and the normal direction are projected onto the XZ plane. FIG. 13 is a diagram showing the relationship between the traveling direction of light emitted from an illumination unit 10c and the normal direction of a point P when the traveling direction and the normal direction are projected onto the YZ plane. FIG. FIG. 13 is a diagram showing the relationship between Ny, φx, and φy when the gloss level α is 5. FIG. 13 is a diagram showing the relationship between Ny, φx, and φy when the gloss level α is 10. FIG. 13 is a diagram showing the relationship between Ny, φx, and φy when the gloss level α is 20. FIG. 13 is a diagram showing the relationship between albedo μ, φx, and φy when gloss level α=5. FIG. 13 is a diagram showing the relationship between albedo μ, φx, and φy when gloss level α=10. FIG. 13 is a diagram showing the relationship between albedo μ, φx, and φy when gloss level α=20. FIG. 13 illustrates the processing performed on the X-direction normal image to generate a shape image. FIG. 13 illustrates the processing performed on a Y-direction normal image to generate a shape image. 1A to 1C are diagrams showing a polarized image obtained by capturing an image of a button battery, and various images generated from the polarized image. 1A to 1C are diagrams showing a polarized image obtained by capturing an image of the side surface of a dry cell, and various images generated from the polarized image. 1A to 1C are diagrams showing a polarized image obtained by capturing an image of the bottom surface of a dry cell, and various images generated from the polarized image. 1A to 1C are diagrams showing a polarized image obtained by capturing an image of a textured resin surface, and various images generated from the polarized image. FIG. 13 is a diagram showing an arrangement of a plurality of illumination units according to a second inspection example. FIG. 2 is a diagram showing a configuration of an illumination unit 10e. FIG. 13 is a diagram showing an example of a plurality of polarized images obtained in a second inspection example. FIG. 13 is a diagram showing an example of the arrangement of a plurality of illumination units according to a third inspection example. 1 is a diagram showing the amount of light that is irradiated from illumination units 10e, 10g to 10i and specularly reflected by an object W and that is transmitted through each polarizer. FIG. 13 is a diagram showing another example of the arrangement of a plurality of illumination units according to the third inspection example. 1 is a diagram showing the amount of light that is irradiated from illumination units 10e, 10g to 10l and specularly reflected by an object W and that is transmitted through each polarizer. 27A to 27D are diagrams showing examples of polarized images 50a to 50d obtained using the multiple illumination units shown in FIG. 26. 29 is a diagram showing an example of polarized images 50a to 50d obtained using the multiple illumination units shown in FIG. 28. FIG. 13 is a diagram showing the arrangement of a plurality of illumination units according to a fourth inspection example. FIG. 13 is a diagram showing an example of a plurality of polarized images obtained in a fourth inspection example. FIG. 13 is a diagram showing the arrangement of a plurality of illumination units according to a fifth inspection example. FIG. 13 is a diagram showing an example of a plurality of polarized images obtained in the fifth inspection example.

The embodiment of the present invention will be described in detail with reference to the drawings. Note that the same or corresponding parts in the drawings will be given the same reference numerals and their description will not be repeated.

§1 Application Example An application example of the present invention will be described with reference to Figures 1 and 2. Figure 1 is a schematic diagram showing the overall configuration of an inspection system 1 according to the present embodiment. The inspection system 1 inspects an object W using an image obtained by imaging the object W. The object W has a glossy surface such as metal or glass, i.e., a surface that mainly specularly reflects incident light. The inspection system 1 is incorporated into, for example, a production line, and inspects the object W for defects while it is moving on a conveyor belt 2.

Defects include scratches, unevenness, dirt, dust, burrs, and chips. In order to accurately inspect for defects, it is necessary to make the defective parts stand out in the image. The lighting conditions that make the defective parts stand out differ depending on the type of defect. Therefore, when inspecting for multiple types of defects, multiple images taken under multiple lighting conditions are required. Alternatively, defective parts may stand out in an image obtained by combining multiple images taken under multiple lighting conditions. When inspecting such defects, multiple images taken under multiple lighting conditions are also required. The inspection system 1 shown in FIG. 1 captures multiple images corresponding to multiple lighting conditions for a moving object W in a single image (one-shot image capture), and inspects the object W using the multiple images obtained.

As shown in FIG. 1, the inspection system 1 includes a plurality of illumination units 10 that illuminate an object W, a polarization camera 20, and an inspection device 30.

The multiple illumination units 10 emit illumination light with different polarization states. The inspection system 1 shown in FIG. 1 includes illumination units 10a, 10b, 10c, 10d, .... Hereinafter, when there is no particular distinction between the illumination units 10a, 10b, 10c, 10d, ..., each of the illumination units 10a, 10b, 10c, 10d, ... will be referred to as "illumination unit 10."

Typically, each of the multiple illumination units 10 includes a light-emitting unit that emits unpolarized light and a linear polarizing filter that is disposed between the light-emitting unit and the object W. The polarization direction of the light that passes through the linear polarizing filter differs among the multiple illumination units 10. Note that one of the multiple illumination units 10 may not include a linear polarizing filter. In this case, one illumination unit 10 irradiates the object W with unpolarized light.

The polarization camera 20 has multiple light detection sensors arranged in a matrix, and generates image data (hereinafter simply referred to as "image") in which the amount of light received (brightness) detected by the light detection sensors is the pixel value. The multiple light detection sensors are the same size. The light detection sensors are, for example, CCD (Coupled Charged Device) or CMOS (Complementary Metal Oxide Semiconductor) sensors. A polarizer is provided on the light incident side of each light detection sensor.

Figure 2 is a diagram showing an example of the arrangement of polarizers in the polarization camera 20. As shown in Figure 2, in the polarization camera 20, unit areas 21 including multiple polarizers 22 are repeatedly arranged. The multiple unit areas 21 are arranged in a matrix. The multiple polarizers 22 have the same size.

In the polarization camera 20 illustrated in FIG. 2, the unit area 21 includes, as the multiple polarizers 22, four polarizers 22a-22d that overlap with the four light detection sensors, respectively. Hereinafter, when there is no particular distinction between the polarizers 22a-22d, each of the polarizers 22a-22d will be referred to as "polarizer 22."

Each of the polarizers 22a to 22d transmits linearly polarized light in a predetermined polarization direction. If the polarization direction of the polarizer 22a is the reference direction and the angle between the reference direction and the polarization direction is the polarization angle, the polarization angles of the polarizers 22a to 22d are 0°, 45°, 90°, and 135°, respectively. Note that the polarization angle may include manufacturing errors, installation errors, and the like. For example, if the maximum sum of the manufacturing errors and installation errors is β°, then a "polarization angle of 45°" includes the range of 45°±β°.

The polarization camera 20 captures images with the multiple illumination units 10 turned on simultaneously, and outputs multiple polarization images 50 (see FIG. 1) corresponding to the multiple polarizers 22, respectively. In the inspection system 1 shown in FIG. 1, the multiple polarization images 50 include polarization images 50a, 50b, 50c, 50d, .... Hereinafter, when there is no particular distinction between the polarization images 50a, 50b, 50c, 50d, ..., each of the polarization images 50a, 50b, 50c, 50d, ... will be referred to as "polarized image 50."

The pixel value of each of the multiple polarized images 50 indicates the brightness of the light transmitted through a corresponding one of the multiple polarizers 22. For example, if the polarization camera 20 has the configuration shown in FIG. 2, polarized images 50a, 50b, 50c, and 50d corresponding to the polarizers 22a to 22d, respectively, are output.

The inspection device 30 inspects the object W using multiple polarized images 50 received from the polarization camera 20.

Each of the multiple polarizers 22 transmits more illumination light from the illumination unit 10 that emits linearly polarized light that forms the smallest angle with the polarization direction of the polarizer 22 among the multiple illumination units 10.

For example, a case will be described in which the polarization direction of the polarizer 22a is parallel to the polarization direction of the illumination unit 10a, and the polarization direction of the polarizer 22c is parallel to the polarization direction of the illumination unit 10c. As shown in FIG. 2, the polarization directions of the polarizers 22a and 22c are orthogonal to each other. In this case, the light irradiated from the illumination unit 10a and specularly reflected on the surface of the object W passes through the polarizer 22a but does not pass through the polarizer 22c. On the other hand, the light irradiated from the illumination unit 10c and specularly reflected on the surface of the object W does not pass through the polarizer 22a but passes through the polarizer 22c. Therefore, the polarized image 50a corresponding to the polarizer 22a is captured so that the light of the illumination unit 10a is the strongest. The polarized image 50c corresponding to the polarizer 22c is captured so that the light of the illumination unit 10c is the strongest.

In this way, each of the multiple polarized images 50 is captured so that the light from the illumination unit 10, which emits linearly polarized light that forms the smallest angle with the polarization direction of the corresponding polarizer 22, is the strongest. In other words, multiple polarized images 50 corresponding to multiple lighting conditions are acquired by one-shot imaging. Therefore, a moving object W can be inspected using multiple polarized images 50 with different lighting conditions.

§2 Specific examples <A. Hardware configuration of inspection device>
Fig. 3 is a schematic diagram showing the hardware configuration of the inspection device 30. As shown in Fig. 3, the inspection device 30 typically has a structure conforming to a general-purpose computer architecture, and realizes various processes as described below by executing a program installed in advance by a processor.

More specifically, the inspection device 30 includes a processor 310 such as a CPU (Central Processing Unit) or MPU (Micro-Processing Unit), a RAM (Random Access Memory) 312, a display controller 314, a system controller 316, an I/O (Input Output) controller 318, a hard disk 320, a camera interface 322, an input interface 324, a communication interface 328, and a memory card interface 330. These components are connected to each other so that they can communicate data with each other, with the system controller 316 at the center.

The processor 310 exchanges programs (code) and the like with the system controller 316 and executes them in a predetermined order to achieve the desired computational processing.

The system controller 316 is connected to the processor 310, RAM 312, display controller 314, and I/O controller 318 via buses, and performs data exchange between each component and controls the overall processing of the inspection device 30.

RAM 312 is typically a volatile storage device such as a dynamic random access memory (DRAM), and holds programs read from the hard disk 320, the polarized image 50 acquired by the polarization camera 20, the processing results of the polarized image 50, etc.

The display controller 314 is connected to the display unit 302, and outputs signals for displaying various types of information to the display unit 302 in accordance with internal commands from the system controller 316. Examples of the display unit 302 include a liquid crystal display, an organic EL (Electro Luminescence) display, an organic EL, etc.

The I/O controller 318 controls data exchange between the recording medium and external devices connected to the inspection device 30. More specifically, the I/O controller 318 is connected to the hard disk 320, the camera interface 322, the input interface 324, the communication interface 328, and the memory card interface 330.

The hard disk 320 is typically a non-volatile magnetic storage device, and stores the inspection program 350 executed by the processor 310. The inspection program 350 installed on the hard disk 320 is distributed in a state stored on the memory card 306 or the like. Furthermore, the hard disk 320 stores camera images. Note that instead of the hard disk 320, a semiconductor storage device such as a flash memory or an optical storage device such as a DVD-RAM (Digital Versatile Disk Random Access Memory) may be used.

The camera interface 322 corresponds to an input unit that accepts the polarized image 50 generated by capturing an image of the object W, and mediates data transmission between the processor 310 and the polarization camera 20. More specifically, the camera interface 322 can be connected to the polarization camera 20, and an image capture instruction is output from the processor 310 to the polarization camera 20 via the camera interface 322. As a result, the polarization camera 20 captures the object W, and outputs the generated multiple polarization images 50 to the processor 310 via the camera interface 322.

The input interface 324 mediates data transmission between the processor 310 and an input device such as the keyboard 304, a mouse, a touch panel, or a dedicated console. In other words, the input interface 324 accepts operation commands given by the user operating the input device.

The communication interface 328 mediates data transmission between the processor 310 and other personal computers or server devices (not shown). The communication interface 328 is typically made up of Ethernet (registered trademark) or USB (Universal Serial Bus). As will be described later, instead of installing a program stored in the memory card 306 into the inspection device 30, a program downloaded from a distribution server or the like via the communication interface 328 may be installed into the inspection device 30.

The memory card interface 330 mediates data transmission between the processor 310 and the memory card 306, which is a recording medium. That is, the memory card 306 is distributed in a state in which the inspection program 350 executed by the inspection device 30 and the like are stored, and the memory card interface 330 reads the inspection program 350 from the memory card 306. The memory card 306 is made up of a general-purpose semiconductor storage device such as an SD (Secure Digital), a magnetic recording medium such as a flexible disk, an optical recording medium such as a CD-ROM (Compact Disk Read Only Memory), etc.

<B. Test Examples>
As described above, the optimal illumination conditions differ depending on the type of defect to be inspected. Therefore, the positions, illumination intensity, polarization direction, and the like of the multiple illumination units 10 are appropriately set depending on the type of defect to be inspected. Below, first to fifth inspection examples using the inspection system 1 according to the present embodiment will be described. Note that the inspection system 1 is not limited to the following first to fifth inspection examples, and may be used for other inspection examples.

(B-1. First Inspection Example)
The first inspection example uses a photometric stereo method (photometric stereo method), which is a method for estimating the inclination (normal vector) of the surface of an object by using multiple images obtained under conditions in which the incidence directions of illumination light are different.

Figure 4 shows an example of a conventional lighting device used in photometric stereo. As shown in Figure 4, four arc regions 110a to 110d of a ring-shaped lighting device 110 are sequentially lit and an image is captured using a non-polarized camera. The center of the lighting device 110 is located on the optical axis 225 of the non-polarized camera.

Specifically, the first image is captured with only the arc region 110a illuminated. Next, the second image is captured with only the arc region 110b illuminated. Next, the third image is captured with only the arc region 110c illuminated. Next, the fourth image is captured with only the arc region 110d illuminated. This results in four images captured under conditions with different directions of incidence of the illumination light. However, because four images are captured, the illumination device 110 shown in FIG. 4 cannot be applied to a moving object W. Therefore, in the first inspection example, the inspection system 1 according to this embodiment captures multiple images used in the photometric stereo method in one shot.

(Arrangement of lighting units)
Fig. 5 is a diagram showing the arrangement of a plurality of illumination units according to inspection example 1. As shown in Fig. 5, the plurality of illumination units 10 include illumination units 10a to 10d.

Illumination units 10a and 10c are positioned symmetrically with respect to the optical axis 25 of the polarization camera 20. Illumination units 10b and 10d are positioned symmetrically with respect to the optical axis 25 of the polarization camera 20. Around the optical axis 25 of the polarization camera 20, the difference between the azimuth angle at which illumination unit 10a is positioned and the azimuth angle at which illumination unit 10b is positioned is 90°. Hereinafter, the azimuth direction in which illumination unit 10a is positioned with respect to the optical axis 25 of the polarization camera 20 is defined as the X direction, and the azimuth direction in which illumination unit 10b is positioned with respect to the optical axis 25 of the polarization camera 20 is defined as the Y direction.

The illumination units 10a to 10d each have a light-emitting unit 11a to 11d and a linear polarizing filter 12a to 12d.

The light-emitting units 11a to 11d emit non-polarized light. Each of the light-emitting units 11a to 11d is arc-shaped with a central angle of 90°. The arc-shaped light-emitting units 11a to 11d are arranged so that their centers coincide and they do not overlap with each other. Specifically, one end of each of the light-emitting units 11a to 11d contacts the other end of each of the light-emitting units 11b to 11d and 11a. That is, the combination of the light-emitting units 11a to 11d forms one ring-shaped light. In other words, the light-emitting units 11a to 11d are four arc-shaped regions with a central angle of 90° of one ring-shaped light. The centers of the arc-shaped light-emitting units 11a to 11d are located on the optical axis 25 of the polarization camera 20.

Linear polarizing filters 12a to 12d are attached to the light-emitting surfaces of light-emitting units 11a to 11d, respectively.

Figure 6 is a diagram showing an example of the polarization direction of linear polarizing filters 12a to 12d. If the polarization direction of light passing through linear polarizing filter 12a is taken as the reference direction and the angle between the reference direction and the polarization direction is taken as the polarization angle, then the polarization angles of linear polarizing filters 12a to 12d are 0°, 45°, 90°, and 135°, respectively. Note that the polarization angle may include manufacturing errors, installation errors, and the like. For example, if the maximum sum of manufacturing errors and installation errors is β°, then a "polarization angle of 45°" includes the range of 45°±β°. Thus, the angles between the polarization direction of the illumination light of illumination units 10b to 10d and the polarization direction of the illumination light of illumination unit 10a are 45°, 90°, and 135°, respectively.

In the first inspection example, the inspection system 1 includes a polarization camera 20 including polarizers 22a to 22d shown in FIG. 2. The polarization directions of the polarizers 22a to 22d are parallel to the polarization directions of the linear polarization filters 12a to 12d, respectively. Therefore, the light irradiated from the illumination units 10a to 10d and specularly reflected by the object W passes through the polarizers 22a to 22d, respectively. The light irradiated from the illumination unit 10a has components parallel to the polarization directions of the polarizers 22b and 22d. Therefore, some components of the light irradiated from the illumination unit 10a and specularly reflected by the object W pass through the polarizers 22b and 22d. Specifically, the amount of light irradiated from the illumination unit 10a and passing through each of the polarizers 22b and 22d is approximately 1/2 of the amount of light irradiated from the illumination unit 10a and passing through the polarizer 22a. Similarly, some components of the light irradiated from the illumination unit 10b and specularly reflected by the object W are transmitted through the polarizers 22a and 22c. Some components of the light irradiated from the illumination unit 10c and specularly reflected by the object W are transmitted through the polarizers 22b and 22d. Some components of the light irradiated from the illumination unit 10d and specularly reflected by the object W are transmitted through the polarizers 22a and 22c.

(Inspection process flow)
7 is a flow chart showing an example of the flow of the inspection process in the first inspection example. First, the polarization camera 20 captures an image of the object W while the illumination units 10a to 10d are simultaneously turned on, and outputs the polarization images 50a to 50d corresponding to the polarizers 22a to 22d, respectively (step S1). After step S1, the inspection system 1 starts steps S2, S3, and S10 in parallel.

In step S2, the processor 310 of the inspection device 30 uses the polarized images 50a to 50d to generate an X-direction normal image that indicates the magnitude of the component along the X direction in the normal vector of the surface of the object W. Similarly, in step S3, the processor 310 uses the polarized images 50a to 50d to generate a Y-direction normal image that indicates the magnitude of the Y-direction component in the normal vector of the surface of the object W. Details of the X-direction normal image and the method of generating the X-direction normal image will be described later.

After step S2 and step S3, step S4 and step S5 are respectively performed. In steps S2 and S3, as described below, an X-direction normal image and a Y-direction normal image are generated on the assumption that the glossiness α of the surface of the object W is 1. However, the glossiness α of the surface of the object W is not necessarily 1. The proportion of the component of the incident light that is specularly reflected differs depending on the glossiness α. Therefore, in step S4, the processor 310 corrects the X-direction normal image according to the glossiness α of the surface of the object W. In step S5, the processor 310 corrects the Y-direction normal image according to the glossiness α of the surface of the object W. The method of correcting the X-direction normal image and the Y-direction normal image will be described in detail below.

After steps S4 and S5, the processor 310 starts steps S6 and S8 in parallel.

In step S6, the processor 310 generates a shape image showing the shape of the surface of the object W based on the X-direction normal image and the Y-direction normal image. The method of generating the shape image will be described in detail later.

Next, in step S7, the processor 310 inspects the surface of the object W based on the shape image. For example, the processor 310 inspects for defects that cause unevenness, such as scratches and dents. In addition, in step S7, the processor 310 may generate a binary image by binarizing the shape image, and inspect for the presence or absence of defects based on the binary image.

In step S8, the processor 310 generates an albedo image that indicates the proportion of light that is specularly reflected out of the incident light (albedo (reflectance)). The albedo image is generated using the polarized images 50a to 50d and the X-direction normal image and the Y-direction normal image corrected in steps S4 and S5, respectively. The method of generating the albedo image will be described in detail later.

Next, in step S9, processor 310 inspects the surface of object W based on the albedo image. For example, processor 310 inspects for the presence or absence of defects that cause a change in the albedo, such as dirt. Note that in step S9, processor 310 may generate a binary image by binarizing the albedo image, and inspect for the presence or absence of defects based on the binary image.

In step S10, the processor 310 generates an average image by averaging the polarized images 50a-50d. The value of each pixel in the average image is the average value of the values of that pixel in the polarized images 50a-50d. The average image corresponds to an image obtained by simultaneously turning on the light-emitting units 11a-11d with the linear polarizing filters 12a-12d removed, and capturing an image using a non-polarized camera instead of the polarized camera 20.

Next, in step S11, the processor 310 uses the average image to determine the position of the object W, etc.

After steps S7, S9, and S11 are completed, the processor 310 causes the display unit 302 to display the test results (step S12). After step S12, the processor 310 ends the test process.

(X-direction normal image and method for generating the X-direction normal image)
Fig. 8 is a diagram showing the relationship between the illumination units 10a to 10d and the normal vector n at point P on the surface of the object W. In Fig. 8, the optical axis 25 of the polarization camera 20 coincides with the Z axis. The normal vector n is expressed as (nx, ny, nz), where nx, ny, and nz are the X, Y, and Z components of the normal vector n, respectively.

Fig. 9 is a diagram showing the relationship between the traveling direction of light emitted from the illumination unit 10c and the normal direction of point P when the traveling direction and the normal direction are projected onto the XZ plane. In Fig. 9, the normal direction is represented by a vector (nx, nz). If the angle between the normal direction and the Z axis is φx, φx is represented by the following formula.
φx=arctan(nx/nz)
Furthermore, if the angle of incidence of the light irradiated from each of the illumination units 10a to 10d is θ, nx>0 and nz>0, then the angle between the Z axis and the projection component onto the XZ plane of the direction of travel (direction of specular reflection) of the light irradiated from the illumination unit 10c and specularly reflected at point P is θ+2φx.

Fig. 10 is a diagram showing the relationship between the traveling direction of light emitted from the illumination unit 10c and the normal direction of point P when the traveling direction and the normal direction are projected onto the YZ plane. In Fig. 10, the normal direction is represented by a vector (ny, nz). If the angle between the normal direction and the Z axis is φy, φy is represented by the following formula.
φx=arctan(ny/nz)
Furthermore, if the angle of incidence of the light irradiated from the illumination unit 10c is θ, ny>0 and nz>0, the angle between the Z axis and the projection component of the direction of travel (direction of specular reflection) of the light irradiated from the illumination unit 10c and specularly reflected at point P onto the YZ plane is 2φy.

Similarly, the angles that the projection component of the traveling direction (specular reflection direction) of the light irradiated from illumination units 10a, 10b, and 10d and specularly reflected at point P makes with the Z axis and the projection component onto the XZ plane are θ-2φx, 2φx, and 2φx, respectively. The angles that the projection component of the traveling direction (specular reflection direction) of the light irradiated from illumination units 10a, 10b, and 10d and specularly reflected at point P makes with the Z axis and the projection component onto the YZ plane are 2φy, θ-2φy, and θ+2φy, respectively.

Of the light emitted from the illumination units 10a to 10d and specularly reflected at point P, the intensities Ia to Id of the light that enters the polarization camera 20 are approximated by the following equations using Phong's reflection model.
Ia=μ[cos ^α (θ−2φx)cos ^α (2φy)]
Ib=μ[cos ^α (2φx)cos ^α (θ−2φy)]
Ic=μ[cos ^α (θ+2φx) cos ^α (2φy)]
Id=μ[cos ^α (2φx)cos ^α (θ+2φy)]
In the above formula, α represents the glossiness, and μ represents the albedo (reflectance) of the surface of the object W.

As described above, the polarization direction of the polarizer 22a of the polarization camera 20 is parallel to the polarization direction of the illumination unit 10a and perpendicular to the polarization direction of the illumination unit 10c. Furthermore, the angle between the polarization direction of the polarizer 22a and the polarization directions of the illumination units 10b and 10d is 45°. Therefore, if the pixel value of the polarized image 50a corresponding to the polarizer 22a is Ja,
Ja=Ia+(Ib+Id)/2
Similarly, pixel values Jb to Jd of the polarized images 50b to 50d are expressed as follows:
Jb=Ib+(Ia+Ic)/2
Jc=Ic+(Id+Ib)/2
Jd=Id+(Ic+Ia)/2
It is expressed as:

When the gloss level α=1, the following equations (1) to (3) hold true.
Ja+Jb+Jc+Jd=2(Ia+Ib+Ic+Id)
=8μcos(θ)cos(2φx)cos(2φy)...Formula (1)
Ja-Jc=Ia-Ic=2μsin(θ)sin(2φx)cos(2φy)...Formula (2)
Jb-Jd=Ib-Id=2μsin(θ)cos(2φx)sin(2φy)...Equation (3).

The following formula (4) is derived from the above formulas (1) and (2): Similarly, the following formula (5) is derived from the above formulas (1) and (3):
tan(2φx)={4(Ja-Jc)}/{(Ja+Jb+Jc+Jd)tan(θ)}
...Formula (4)
tan(2φy)={4(Jb-Jd)}/{(Ja+Jb+Jc+Jd)tan(θ)}
...Formula (5).

In formula (4) and formula (5), the incident angle θ is determined in advance according to the positions of the illumination units 10a to 10d. Therefore, the processor 310 calculates the right side of formula (4) using the pixel values of the polarized images 50b to 50d and the incident angle θ, and generates an X-direction normal image with the pixel value Nx as the result. Similarly, the processor 310 calculates the right side of formula (5) using the pixel values of the polarized images 50b to 50d and the incident angle θ, and generates a Y-direction normal image with the pixel value Ny as the result. tan(2φx)(≡Nx) expressed in formula (4) is a value that depends on nx, which is the x-direction component of the normal vector. tan(2φx)(≡Ny) expressed in formula (5) is a value that depends on ny, which is the y-direction component of the normal vector. Therefore, the X-direction normal image and the Y-direction normal image represent the normal direction of the surface of the object W.

(Method of correcting X-direction normal image and Y-direction normal image)
The X-direction normal image and the Y-direction normal image are generated according to formulas (4) and (5), respectively, assuming that the glossiness α of the surface of the object W is 1. Therefore, the X-direction normal image and the Y-direction normal image are corrected according to the glossiness α of the object W. As a result, a normal image according to the glossiness α is generated with high accuracy.

The gloss level α of the object W is determined in advance according to the following pre-inspection procedure. In the pre-inspection procedure, the defect-free object W is placed on a two-axis goniostage. The two-axis goniostage is a stage that can be tilted in the X and Y directions.

When the tilt angles of the two-axis goniostage in the X and Y directions are set to 0°, the target object W is placed on the two-axis goniostage so that it has a horizontal, flat area on the top surface (hereinafter referred to as the "target area"). Therefore, the angle φx between the X-direction component of the normal vector of the target area and the Z-axis coincides with the tilt angle of the two-axis goniostage in the X direction. Similarly, the angle φy between the Y-direction component of the normal vector of the target area and the Z-axis coincides with the tilt angle of the two-axis goniostage in the Y direction.

The inspection system 1 is used to capture M images while changing the tilt angles of the two-axis goniostage in the X and Y directions. For each of the M images, an X-direction normal image and a Y-direction normal image are generated. The processor 310 stores a data set [φxm, φym, Nxm, Nym] in the RAM 312 for any pixel in the target area in the X-direction normal image and the Y-direction normal image. m indicates the image number and is an integer between 1 and M. φxm is the X-direction tilt angle of the two-axis goniostage in the mth image capture. φym is the Y-direction tilt angle of the two-axis goniostage in the mth image capture. Nxm is the pixel value of the X-direction normal image. Nym is the pixel value of the Y-direction normal image.

The processor 310 determines the gloss level α that best fits the data set [φxm, φym, Nxm, Nym] stored in the RAM 312 using the following theoretical formula:
≪Theoretical formula≫
Ia=cos ^α (θ−2φxm) cos ^α (2φym)
Ib=cos ^α (2φxm) cos ^α (θ−2φym)
Ic=cos ^α (θ+2φxm) cos ^α (2φym)
Id=cos ^α (2φxm) cos ^α (θ+2φym)
Nx={2(Ia-Ic)}/{(Ia+Ib+Ic+Id)tan(θ)}
Ny={2(Ib-Id)}/{(Ia+Ib+Ic+Id)tan(θ)}
For example, the processor 310 determines the gloss value α using a nonlinear least squares method. By such a pre-inspection procedure, the gloss value α of the object W is determined in advance.

Figure 11 is a diagram showing the relationship between Ny, φx, and φy when gloss level α = 5. Figure 12 is a diagram showing the relationship between Ny, φx, and φy when gloss level α = 10. Figure 13 is a diagram showing the relationship between Ny, φx, and φy when gloss level α = 20.

As shown in Figures 11 to 13, the relationship between the angles φx, φy between the X- and Y-direction components of the normal vector and the Z-axis, respectively, and Ny, which is the calculation result of the right-hand side of equation (5), shows a monotonous change. That is, Ny increases as φy increases. Furthermore, if φy>0, Ny decreases as φx increases. If φy<0, Ny increases as φx increases. However, while the relationship between Ny and φy is linear when the gloss level α is close to 1, the relationship between Ny and φy becomes nonlinear as the gloss level α increases. Therefore, the processor 310 corrects the normal image according to the gloss level α.

Specifically, processor 310 uses the above theoretical formula into which a predetermined gloss level α is substituted to back-calculate (φx, φy) from (Nx, Ny). Processor 310 corrects the X-direction normal image by replacing each pixel value Nx in the X-direction normal image with φx. Similarly, processor 310 corrects the Y-direction normal image by replacing each pixel value Ny in the Y-direction normal image with φy.

Alternatively, the processor 310 may select one lookup table corresponding to the gloss level α from a group of pre-created lookup tables, and convert (Nx, Ny) to (φx, φy) using the selected lookup table. The lookup table is created in advance using the above theoretical formula into which the corresponding gloss level α is substituted. By using the lookup table, the time required for the correction process of the normal image is reduced.

(Method of generating albedo image)
When the gloss level α of the surface of the object W is 1, the following equation (6) holds true.
μ={(Ja+Jb+Jc+Jd)/2}[{1+tan ² (2φx)}{1+tan ² (2φy)}] ^1/2 /{4cos(θ)}...Equation (6).

Figure 14 shows the relationship between the albedo μ, φx, and φy when the gloss level α = 5. Figure 15 shows the relationship between the albedo μ, φx, and φy when the gloss level α = 10. Figure 16 shows the relationship between the albedo μ, φx, and φy when the gloss level α = 20.

As shown in Figures 14 to 16, the relationship between the albedo μ and φx and φy changes monotonically. However, when the glossiness α is close to 1, the relationship between the albedo μ and φx and φy is linear, whereas as the glossiness α increases, the relationship between the albedo μ and φx and φy becomes nonlinear. Therefore, the processor 310 generates an albedo image by substituting the pixel values φx and φy of the X-direction normal image and the Y-direction normal image corrected according to the glossiness α, and the pixel values Ja to Jd of the polarized images 50a to 50d, respectively, into the above formula. Each pixel value of the albedo image indicates the albedo μ.

(Method of generating shape image)
The shape image is generated based on the X-direction normal image and the Y-direction normal image corrected according to the gloss level α.

Figure 17 shows the processing performed on the X-direction normal image to generate a shape image. The pixel values of the X-direction normal image indicate the magnitude of the X-direction component of the normal vector of the surface of the object W.

The processor 310 selects one point surrounded by four pixels in the X-direction normal image as the attention point Q. The horizontal direction of the X-direction normal image corresponds to the X direction. Therefore, as shown in FIG. 17, the processor 310 sets rectangular regions R1 and R2 on the left and right sides of the attention point Q, respectively. The height of the rectangular regions R1 and R2 is a predetermined length L, and the width of the rectangular regions R1 and R2 is L/2.

The processor 310 calculates the sum Sx1 of the pixel values contained in the rectangular region R1. Similarly, the processor 310 calculates the sum Sx2 of the pixel values contained in the rectangular region R2. The processor 310 calculates the difference Sx between the sum Sx1 and the sum Sx2. The processor 310 calculates the difference Sx for all points surrounded by four pixels in the X-direction normal image.

Figure 18 shows the processing performed on the Y-direction normal image to generate a shape image. The pixel values of the Y-direction normal image indicate the magnitude of the Y-direction component of the normal vector of the surface of the object W.

The processor 310 selects one point surrounded by four pixels in the Y direction normal image as the attention point Q. The vertical direction of the Y direction normal image corresponds to the Y direction. Therefore, as shown in FIG. 18, the processor 310 sets rectangular regions R3 and R4 above and below the attention point Q, respectively. The width of the rectangular regions R3 and R4 is L, and the height of the rectangular regions R3 and R4 is L/2.

The processor 310 calculates Sy1, the value of the pixel contained in the rectangular region R3. Similarly, the processor 310 calculates Sy2, the value of the pixel contained in the rectangular region R4. The processor 310 calculates the difference Sy between the sum Sy1 and the sum Sy2. The processor 310 calculates the difference Sy for all points surrounded by four pixels in the Y direction normal image.

When a defect such as a scratch or a dent is present on the surface of the object W, the normal vector changes significantly at the defect. The greater the change in the X-direction component of the normal vector, the greater the difference Sx. The greater the change in the Y-direction component of the normal vector, the greater the difference Sy. In other words, the differences Sx and Sy take large values in locations where there are defects that cause unevenness, such as scratches or dents. Therefore, the processor 310 generates a shape image in which S, obtained by the following formula, is the pixel value.
S=A(Sx+Sy)+B
A is a parameter for determining the contrast of the shape image. B is a parameter for determining the overall pixel value level of the shape image. The processor 310 sets the values of parameters A and B so that the maximum, minimum, and difference between the maximum and minimum pixel values in the shape image fall within a specified range. The specified range is determined in advance according to the contrast and level suitable for defect inspection. In this way, a shape image suitable for inspection of defects that cause unevenness is generated.

(Examples of polarization image, normal image, shape image, albedo image, and average image)
19 is a diagram showing polarized images obtained by capturing an image of a button battery, and various images generated from the polarized images. The processor 310 generates an X-direction normal image 51 and a Y-direction normal image 52 from the polarized images 50a to 50d. The processor 310 generates a shape image 53 based on the X-direction normal image 51 and the Y-direction normal image 52. The processor 310 binarizes the shape image 53 to generate a binary image 54. Furthermore, the processor 310 generates an albedo image 55 and an average image 56 from the polarized images 50a to 50d.

In FIG. 19, a dent is present in the area surrounded by frame line F1 on the surface of the button battery. Furthermore, dirt is present in the area surrounded by frame line F2 on the surface of the button battery. As shown in FIG. 19, in shape image 53, a change in pixel value corresponding to the dent portion is observed within frame line F1. In binary image 54, the dent portion is represented in white. As a result, by using shape image 53 or binary image 54, defects that cause unevenness, such as scratches or dents, can be inspected with high accuracy. Note that in X-direction normal image 51 and Y-direction normal image 52, pixel values are different between the dent portion and its surroundings. Therefore, defects that cause unevenness can also be inspected using X-direction normal image 51 and Y-direction normal image 52.

In the X-direction normal image 51, the Y-direction normal image 52, the shape image 53, and the binary image 54, no change in pixel value is observed between the dirty area and its surroundings. This is because the dirt has not caused any change in the unevenness. In contrast, in the albedo image 55, a change in pixel value corresponding to the dirt is observed within the frame line F2. This is because the dirt has caused a change in the albedo (reflectance). In this way, by using the albedo image 55, defects that cause changes in the albedo (reflectance), such as dirt, can be accurately inspected.

Figure 20 shows a polarized image obtained by capturing an image of the side of a dry cell, and various images generated from the polarized image. Figure 20 shows polarized images 50a to 50d, an X-direction normal image 51, a Y-direction normal image 52, and an average image 56 generated from the polarized images 50a to 50d, and a shape image 53 and a binary image 54 generated from the X-direction normal image 51 and the Y-direction normal image 52. In Figure 20, a dent is present in the area surrounded by a frame line F1 on the side of the dry cell. In the shape image 53, a change in pixel value corresponding to the dent portion is observed within the frame line F1. In the binary image 54, the dent portion is displayed in white. As a result, by using the shape image 53 or the binary image 54, defects that cause unevenness, such as scratches or dents, can be inspected with high accuracy.

Figure 21 shows a polarized image obtained by capturing an image of the bottom surface of a dry cell, and various images generated from the polarized images. Figure 21 shows polarized images 50a to 50d, an X-direction normal image 51, a Y-direction normal image 52, and an average image 56 generated from the polarized images 50a to 50d, and a shape image 53 generated from the X-direction normal image 51 and the Y-direction normal image 52. In Figure 21, a dent is present in the area surrounded by a frame line F1 on the bottom surface of the dry cell. In the shape image 53, a change in pixel value corresponding to the dent portion is observed within the frame line F1. As a result, by using the shape image 53, defects that cause unevenness, such as scratches or dents, can be accurately inspected.

Figure 22 shows polarized images obtained by capturing an image of a textured resin surface, and various images generated from the polarized images. Figure 22 shows polarized images 50a to 50d, an X-direction normal image 51, a Y-direction normal image 52, and an average image 56 generated from the polarized images 50a to 50d, and a shape image 53 and a binary image 54 generated from the X-direction normal image 51 and the Y-direction normal image 52. In Figure 22, a dent is present in the area surrounded by frame line F1 on the resin surface. Furthermore, dirt is present in the area surrounded by frame line F2 on the resin surface.

As shown in FIG. 22, in shape image 53, a change in pixel value corresponding to the dent portion is observed within frame line F1. In binary image 54, the dent portion is shown in white. As a result, by using shape image 53 or binary image 54, defects that cause unevenness, such as scratches or dents, can be inspected with high accuracy.

In the X-direction normal image 51, the Y-direction normal image 52, the shape image 53, and the binary image 54, no change in pixel value is observed between the dirty area and its surroundings. This is because the dirt has not caused any change in the unevenness. In contrast, in the average image 56, a change in pixel value corresponding to the dirty area is observed within the frame line F2. This is because the dirt has increased the albedo (reflectance). In this way, even when the average image 56 is used, defects that cause changes in albedo (reflectance), such as dirt, can be accurately inspected.

(B-2. Second Inspection Example)
The second inspection example uses two images obtained under bright-field and dark-field illumination conditions. Bright-field illumination is an illumination method in which specularly reflected light from the surface of the object W is incident on the polarization camera 20. Dark-field illumination is an illumination method in which specularly reflected light from the surface of the object W is not incident on the polarization camera 20, and only diffusely reflected light is incident on the polarization camera 20. Light is not easily specularly reflected in the area where the scratch exists. Therefore, in the image obtained under the bright-field illumination condition, the brightness of the area where the scratch exists is reduced. As a result, the image obtained under the bright-field illumination condition can be used to inspect the scratch with high accuracy. On the other hand, in the area where dirt or foreign matter exists, light is easily diffusely reflected. Therefore, in the image obtained under the dark-field illumination condition, the brightness of the area where dirt or foreign matter exists is high. As a result, the image obtained under the dark-field illumination condition can be used to inspect the dirt or foreign matter with high accuracy.

(Arrangement of lighting units)
Fig. 23 is a diagram showing the arrangement of a plurality of illumination units according to the second inspection example. As shown in Fig. 23, the plurality of illumination units 10 include illumination units 10e and 10f. The illumination units 10e and 10f are arranged so that their elevation angles with respect to the object W are different from each other.

The illumination unit 10f is a ring-shaped illumination unit centered on the optical axis 25 of the polarization camera 20. The illumination unit 10f has a ring-shaped light-emitting unit 11f centered on the optical axis 25 of the polarization camera 20, and a linear polarizing filter 12f attached to the light-emitting surface of the light-emitting unit 11f. The elevation angle of the illumination unit 10f with respect to the object W is set so that specularly reflected light from the object W does not enter the polarization camera 20.

Figure 24 is a diagram showing the configuration of the illumination unit 10e. The illumination unit 10e is a coaxial illumination that irradiates illumination light onto the object W along the optical axis 25 of the polarization camera 20. As shown in Figure 24, the illumination unit 10e has a light-emitting unit 11e, a linear polarizing filter 12e, and a half mirror 13e. The linear polarizing filter 12e is disposed between the light-emitting surface of the light-emitting unit 11e and the half mirror 13e, and transmits linearly polarized light along a predetermined direction among the non-polarized light irradiated from the light-emitting unit 11e. The half mirror 13e reflects the linearly polarized light that has passed through the linear polarizing filter 12e down along the optical axis 25 of the polarization camera 20.

The linear polarizing filters 12e and 12f are positioned so that the polarization directions of the linearly polarized light irradiated from the illumination units 10e and 10f to the object W are perpendicular to each other.

(Testing Method)
In the second inspection example, the inspection system 1 includes a polarization camera 20 including polarizers 22a to 22d shown in FIG. 2. The polarization direction of the polarizer 22a coincides with (is parallel to) the polarization direction of the linearly polarized light irradiated from the illumination unit 10e to the object W. The polarization direction of the polarizer 22c coincides with (is parallel to) the polarization direction of the linearly polarized light irradiated from the illumination unit 10f. Therefore, the light irradiated from the illumination unit 10e and specularly reflected by the object W passes through the polarizer 22a, but does not pass through the polarizer 22c. A part of the light irradiated from the illumination unit 10f and diffusely reflected by the object W passes through the polarizer 22c. Note that the light irradiated from the illumination unit 10f and specularly reflected by the object W does not enter the polarization camera 20.

Figure 25 shows an example of multiple polarized images obtained in the second inspection example. Polarized images 50a to 50d correspond to polarizers 22a to 22d, respectively.

The polarized image 50a shows the brightness of the light that is irradiated from the illumination unit 10e and specularly reflected by the object W. In other words, the polarized image 50a corresponds to an image obtained by imaging under bright field illumination conditions. Therefore, scratches can be easily confirmed within the frame line F1 of the polarized image 50a. In this way, scratches can be inspected with high precision by using the polarized image 50a.

The polarized image 50c shows the brightness of the light that is irradiated from the illumination unit 10f, diffusely reflected by the object W, and transmitted through the polarizer 22c. In other words, the polarized image 50c corresponds to an image obtained by imaging under dark-field illumination conditions. Therefore, dirt can be easily confirmed within the frame line F2 of the polarized image 50c. In this way, the use of the polarized image 50c allows dirt to be inspected with high accuracy.

Polarized images 50b and 50d show intermediate brightness between polarized images 50a and 50c. Therefore, polarized images 50b and 50d do not need to be used to inspect object W.

(B-3. Third Inspection Example)
The third inspection example uses a phase shift method, which is a method for measuring the three-dimensional shape of an object W from a plurality of images captured by irradiating the object with a stripe pattern whose irradiation intensity is modulated into a sine wave while shifting the phase.

(Arrangement of lighting units)
Fig. 26 is a diagram showing an example of the arrangement of a plurality of illumination units according to the third inspection example. As shown in Fig. 26, the plurality of illumination units 10 include illumination units 10e, 10g to 10i. As described in the second inspection example, illumination unit 10e irradiates the object W with illumination light coaxial with the optical axis 25 of the polarization camera 20 (see Fig. 24).

Each of the illumination units 10g to 10i is a ring-shaped illumination. The illumination units 10g to 10i each have a ring-shaped light-emitting unit 11g to 11i centered on the optical axis 25 of the polarization camera 20, and a linear polarizing filter 12g to 12i attached to the light-emitting surface of the light-emitting unit.

When the polarization direction of the light irradiated from the illumination unit 10e is taken as the reference direction, the linear polarizing filters 12e, 12g to 12i are positioned so that the angles between the polarization direction of the light irradiated from the illumination units 10e, 10g to 10i and the reference direction are 0°, 45°, 90°, and 135°, respectively.

The illumination units 10e, 10g to 10i are arranged so that they have different elevation angles relative to the object W. Specifically, the illumination units 10e, 10g to 10i are arranged so that their elevation angles relative to the object W decrease in this order. In other words, the angles of incidence of illumination light from the illumination units 10e, 10g to 10i to the object W increase in this order. The elevation angle of the illumination unit 10e relative to the object W is 90°, and the angle of incidence of illumination light from the illumination unit 10e to the object W is 0°.

In the third inspection example, the inspection system 1 includes a polarization camera 20 including polarizers 22a to 22d shown in FIG. 2. The polarization directions of the polarizers 22a to 22d are parallel to the polarization directions of the light irradiated from the illumination units 10e, 10g to 10i, respectively. Therefore, the light irradiated from the illumination units 10e, 10g to 10i and specularly reflected by the object W passes through the polarizers 22a to 22d, respectively. The light irradiated from the illumination unit 10e has components parallel to the polarization directions of the polarizers 22b and 22d. Therefore, some components of the light irradiated from the illumination unit 10e and specularly reflected by the object W pass through the polarizers 22b and 22d. Specifically, the amount of light irradiated from the illumination unit 10e and transmitted through each of the polarizers 22b and 22d is approximately 1/2 of the amount of light irradiated from the illumination unit 10e and transmitted through the polarizer 22a. Similarly, some components of the light irradiated from illumination unit 10g and specularly reflected by object W are transmitted through polarizers 22a and 22c. Some components of the light irradiated from illumination unit 10h and specularly reflected by object W are transmitted through polarizers 22b and 22d. Some components of the light irradiated from illumination unit 10i and specularly reflected by object W are transmitted through polarizers 22a and 22c. Light irradiated from illumination units 10e, 10g to 10i and specularly reflected by object W is not transmitted through polarizers 22c, 22d, 22a, and 22b.

Figure 27 is a diagram showing the amount of light that is irradiated from illumination units 10e, 10g to 10i and specularly reflected by object W that is transmitted through each polarizer. As shown in Figure 27, the amount of light that is transmitted through each of polarizers 22a to 22d follows a sine wave. That is, polarized images 50a to 50d corresponding to polarizers 22a to 22d, respectively, correspond to images captured under conditions in which light in a concentric stripe pattern is irradiated onto object W. Furthermore, as shown in Figure 27, the phases of the stripe patterns corresponding to polarizers 22a to 22d, respectively, are shifted by π/2.

Figure 28 is a diagram showing another example of the arrangement of multiple illumination units for the third inspection example. The multiple illumination units 10 shown in Figure 28 differ from the multiple illumination units 10 shown in Figure 26 in that they further include illumination units 10j to 10l.

Each of the lighting units 10j to 10l is a ring-shaped light source. The lighting units 10j to 10l each have a ring-shaped light-emitting unit 11j to 11l centered on the optical axis 25 of the polarization camera 20, and a linear polarizing filter 12j to 12l attached to the light-emitting surface of the light-emitting unit 11j to 11l.

When the polarization direction of the light irradiated from the illumination unit 10e is taken as the reference direction, the linear polarizing filters 12j to 12l are positioned so that the angles between the polarization direction of the light irradiated from the illumination units 10j to 10l and the reference direction are 22.5°, 67.5°, and 110.5°, respectively.

The lighting units 10e, 10j, 10g, 10k, 10h, 10l, and 10i are arranged such that the elevation angle decreases in this order with respect to the object W. In other words, the angle of incidence of the illumination light from the lighting units 10e, 10j, 10g, 10k, 10h, 10l, and 10i to the object W increases in this order.

Figure 29 is a diagram showing the amount of light that is irradiated from illumination units 10e, 10g-10l and specularly reflected by object W that is transmitted through each polarizer. As shown in Figure 29, the amount of light that is transmitted through each of polarizers 22a-22d follows a sine wave. That is, polarized images 50a-50d corresponding to polarizers 22a-22d, respectively, correspond to images captured under conditions in which light in a concentric stripe pattern is irradiated onto object W. Furthermore, as shown in Figure 29, the phases of the stripe patterns corresponding to polarizers 22a-22d, respectively, are shifted by π/2.

(Testing Method)
The processor 310 performs an inspection of the object W based on the polarized images 50a-50d. The processor 310 generates a phase image and a specular image based on the polarized images 50a-50d by a phase shift method.

If the two-dimensional coordinates of an image are represented as (x, y) and the values of pixels (x, y) in the polarization images 50a to 50d are represented as Ja(x, y), Jb(x, y), Jc(x, y), and Jd(x, y), respectively, the phase of the pixel is expressed as follows:
Φ(x,y)=arctan[{Jb(x,y)-Jd(x,y)}/{Ja(x,y)-Jc(x,y)}]
The normal angle of the object W in which the stripe pattern can be captured is in the range of 0° to 45°. The phase value Φ(x, y) is a value proportional to the angle between the normal direction of the object W and the optical axis 25 of the polarization camera 20. The processor 310 generates a phase image in which the value of pixel (x, y) is Ip(x, y) = {2π-Φ(x, y)} × 128/π. At this time, the phase image becomes a white pixel (256) when the angle between the normal direction of the object W and the optical axis 25 is 0°, and becomes a black pixel (0) when the angle between the normal direction of the object W and the optical axis 25 is 45°. Therefore, the unevenness of the surface of the object W can be inspected by using the phase image.

Furthermore, the processor 310 determines whether the value of pixel (x,y) is A(x,y)=[{Ja(x,y)-Jc(x,y)} ² +{Jb(x,y)-Jd(x,y)} ² ] ^1/2
A specular reflection image is generated such that: The specular reflection image represents the intensity of light specularly reflected from the surface of the object W. If the object W has a scratch, the light may be diffusely reflected due to the minute shape of the scratch. In this case, the specular reflection component becomes relatively weak at the scratched portion. Therefore, the specular reflection image can be used to inspect the unevenness of the surface of the object W.

Figure 30 is a diagram showing an example of polarized images 50a to 50d obtained using multiple illumination units 10 shown in Figure 26. Figure 31 is a diagram showing an example of polarized images 50a to 50d obtained using multiple illumination units 10 shown in Figure 28. As shown in Figures 30 and 31, a stripe pattern is observed that is specularly reflected from the surface of the object W, and the phase of the stripe pattern is shifted by π/2 in each of the polarized images 50a to 50d. Furthermore, by increasing the number of illumination units 10, the stripe pattern becomes smoother. This improves the resolution in the normal direction to the surface of the object W.

It is also possible to provide only the ring-shaped illumination units 10g-10i (or illumination units 10g-10l) without providing the illumination unit 10e, which is a coaxial illumination unit. In this case, the stripe pattern in the central part of the concentric circles will be missing. By providing the illumination unit 10e, it is possible to reduce the loss of the stripe pattern in the central part of the concentric circles, allowing for more accurate inspection.

(B-4. Fourth Test Example)
The fourth inspection example uses two images obtained under coaxial illumination and backlight conditions. The image obtained under illumination conditions coaxial with the optical axis 25 of the polarization camera 20 shows the brightness of the specularly reflected light on the surface of the object W. As described above, light is not easily reflected specularly in the area where a scratch exists. Therefore, in the image obtained under coaxial illumination, the brightness of the area where a scratch exists is reduced. As a result, the image obtained under coaxial illumination can be used to inspect the scratch with high accuracy. On the other hand, if the object W has a light-shielding property, the brightness of the area corresponding to the object W is 0 in the image obtained under backlight. Therefore, the outer periphery of the object W can be easily confirmed. As a result, the image obtained under backlight can be used to inspect the outer periphery of the object W for burrs or chips with high accuracy.

(Arrangement of lighting units)
Fig. 32 is a diagram showing the arrangement of a plurality of illumination units according to the fourth inspection example. As shown in Fig. 32, the plurality of illumination units 10 include illumination units 10e and 10m. As described in the second inspection example, illumination unit 10e irradiates illumination light along optical axis 25 of polarization camera 20 onto object W (see Fig. 24).

The illumination unit 10m acts as a backlight and is disposed on the opposite side of the object W from the polarized camera 20. In other words, the illumination unit 10m irradiates illumination light from the back side of the object W. The illumination unit 10m has a flat light-emitting unit 11m and a linear polarizing filter 12m attached to the light-emitting surface of the light-emitting unit 11f.

The linear polarizing filter 12e (see FIG. 24) and the linear polarizing filter 12m of the illumination unit 10e are arranged so that the polarization directions of the linearly polarized light irradiated from the illumination units 10e and 10m to the object W are perpendicular to each other.

(Testing Method)
In the fourth inspection example, the inspection system 1 includes a polarization camera 20 including polarizers 22a to 22d shown in FIG. 2. The polarization direction of the polarizer 22a coincides with (is parallel to) the polarization direction of the linearly polarized light irradiated from the illumination unit 10e to the object W. The polarization direction of the polarizer 22c coincides with (is parallel to) the polarization direction of the linearly polarized light irradiated from the illumination unit 10m. Therefore, the light irradiated from the illumination unit 10e and specularly reflected by the object W passes through the polarizer 22a but does not pass through the polarizer 22c. The light irradiated from the illumination unit 10m and traveling around the object W passes through the polarizer 22c but does not pass through the polarizer 22a.

Figure 33 shows an example of multiple polarized images obtained in the fourth inspection example. Polarized images 50a to 50d correspond to polarizers 22a to 22d, respectively.

The polarized image 50a shows the brightness of the light that is irradiated from the illumination unit 10e and specularly reflected by the object W. Therefore, it is easy to confirm that the brightness of the scratched area is reduced within the frame line F1 of the polarized image 50a. Therefore, the processor 310 can use the polarized image 50a to inspect scratches with high accuracy.

The polarized image 50c shows the brightness of the light that is irradiated from the illumination unit 10m and travels around the object W. In other words, the polarized image 50c corresponds to an image obtained under backlight conditions. Therefore, burrs on the outer periphery of the object W can be easily confirmed within the frame line F3 of the polarized image 50c. Therefore, the processor 310 can use the polarized image 50c to accurately inspect for burrs or chips on the outer periphery of the object W.

Polarized images 50b and 50d show intermediate brightness between polarized images 50a and 50c. Therefore, polarized images 50b and 50d do not need to be used to inspect object W.

(B-5. Fifth Test Example)
The fifth inspection example uses a high dynamic range technique, which is a technique for generating an image (high dynamic range image) having a wide dynamic range with little blown out highlights or crushed shadows by synthesizing multiple images taken under conditions of different illumination intensities.

(Arrangement of lighting units)
Fig. 34 is a diagram showing the arrangement of a plurality of illumination units according to inspection example 5. As shown in Fig. 34, the plurality of illumination units 10 include illumination units 10n and 10o.

The illumination unit 10n has a light-emitting unit 11n that emits unpolarized light and a linear polarizing filter 12n that is placed on the object W of the light-emitting unit 11n. Therefore, linear polarized light is irradiated from the illumination unit 10n to the object W.

The illumination unit 10o has a light-emitting unit 11o that emits unpolarized light and does not have a linear polarization filter. Therefore, unpolarized light is irradiated from the illumination unit 10o onto the object W.

It is preferable that the illumination units 10n and 10o are arranged close to each other. This makes it possible to obtain, in a single capture, a plurality of polarized images 50 with different illumination intensities and substantially the same illumination direction as illumination conditions, as described below.

(Testing Method)
In the fifth inspection example, the inspection system 1 includes a polarization camera 20 including polarizers 22a to 22d shown in FIG. 2. The polarization direction of the linearly polarized light irradiated from the illumination unit 10n is parallel to the polarization direction of the polarizer 22a and perpendicular to the polarization direction of the polarizer 22c. Therefore, the light irradiated from the illumination unit 10n and specularly reflected by the object W passes through the polarizer 22a and does not pass through the polarizer 22c. The light irradiated from the illumination unit 10n and specularly reflected by the object W has components parallel to the polarization directions of the polarizers 22b and 22d. Therefore, some components of the light irradiated from the illumination unit 10n and specularly reflected by the object W pass through the polarizers 22b and 22d. Specifically, the amount of light irradiated from the illumination unit 10n and passing through each of the polarizers 22b and 22d is about 1/2 of the amount of light irradiated from the illumination unit 10n and passing through the polarizer 22a.

The unpolarized light emitted from the illumination unit 10o has components that are evenly parallel to the polarization direction of the polarizers 22a to 22d. Therefore, the amount of light emitted from the illumination unit 10o that passes through each of the polarizers 22a to 22d is the same.

The intensities of light irradiated from the illumination units 10n and 10o, specularly reflected by the object W, and transmitted through the polarizer 22a are denoted as In and Io, respectively. In this case, pixel values Ja to Jd of the polarized images 50a to 50d corresponding to the polarizers 22a to 22d, respectively, are expressed as follows:
Pixel value Ja of polarized image 50a: In+Io
Pixel value Jb of polarized image 50b: In/2+Io
Pixel value Jc of polarized image 50c:Io
Pixel value Jd of polarized image 50d: In/2+Io
From the above, the polarized images 50a to 50c correspond to a plurality of images taken under conditions of different illumination intensities. Therefore, the processor 310 performs high dynamic range synthesis on the polarized images 50a to 50c to generate a synthetic image. The processor 310 may generate the synthetic image using a known high dynamic range synthesis method.

Figure 35 is a diagram showing an example of multiple polarized images obtained in the fifth inspection example. As shown in Figure 35, polarized image 50a has the highest brightness, polarized image 50c has the lowest brightness, and polarized images 50b and 50d have intermediate brightness. Processor 310 performs high dynamic range synthesis of polarized images 50a to 50c to generate composite image 57. Composite image 57 has little blown-out highlights or crushed shadows. Therefore, by using composite image 57, processor 310 can inspect object W with high accuracy.

C. Modifications
The number of polarizers 22 included in the unit area 21 of the polarization camera 20 is not limited to four, and may be any number. As described above, in the second and fourth inspection examples, the polarized images 50b and 50d may not be used. Therefore, the unit area 21 may include only the polarizers 22a and 22c, and not the polarizers 22b and 22d. Similarly, in the fifth inspection example, the polarized image 50d is not used. Therefore, the unit area 21 may include only the polarizers 22a to 22c, and not the polarizer 22d.

In the first inspection example, the multiple illumination units 10 may include only illumination units 10a and 10c, and may not include illumination units 10b and 10d. In this case, the unit area 21 of the polarization camera 20 may include only polarizers 22a to 22c, and may not include polarizer 22d. The pixel values Ja and Jc of the polarization images 50a and 50c, respectively, are expressed by the following equations.
Ja=Ia=μ[cos ^α (θ−2φx)cos ^α (2φy)]
Jc=Ic=μ[cos ^α (θ+2φx) cos ^α (2φy)]
When the surface glossiness α of the object W is 1, the following equations (7) and (8) hold.
Ja+Jc=2μcos(θ)cos(2φx)cos(2φy)...Formula (7)
Ja-Jc=2μsin(θ)sin(2φx)cos(2φy)...Equation (8).

From the above equations (7) and (8), the following equation (9) is derived.
tan(2φx)=(Ja-Jc)/{(Ja+Jc)tan(θ)}...Equation (9).

In equation (9), the incidence angle θ is determined in advance according to the positions of the illumination units 10a and 10c. Therefore, the processor 310 calculates the right side of equation (9) using the pixel values Ja and Jc of the polarized images 50a and 50c and the incidence angle θ, and generates an X-direction normal image with the resulting pixel value Nx. The value tan(2φx) (≡Nx) expressed in equation (9) is a value that depends on nx, which is the x-direction component of the normal vector. Therefore, the X-direction normal image represents the normal direction of the surface of the object W.

Furthermore, as shown in FIG. 17, the processor 310 sets rectangular regions R1 and R2 to the left and right of the attention point Q, respectively. Then, the processor 310 calculates the difference Sx between the sum Sx1 of the pixel values contained in the rectangular region R1 and the sum Sx2 of the pixel values contained in the rectangular region R2. The processor 310 calculates the difference Sx for all points surrounded by four pixels in the X-direction normal image. The processor 310 can generate a shape image with the difference Sx as the pixel value. As described above, the difference Sx has a large value in a location where there is a defect that causes unevenness, such as a scratch or a dent. Therefore, the processor 310 can accurately inspect defects that cause unevenness by using the shape image.

In the first inspection example, the X-direction normal image and the Y-direction normal image are generated using the above formulas (4) and (5), respectively. However, the method of generating the normal images is not limited to this. For example, a learning model is constructed using a deep learning image generation technique to estimate a normal image from multiple polarization images. The processor 310 may use the learning model to generate a normal image from multiple polarization images.

§3 Supplementary Note As described above, the present embodiment includes the following disclosure.

(Configuration 1)
An inspection system (1), comprising:
A plurality of illumination units (10, 10a to 10o) for illuminating an object (W);
A polarization camera (20) in which unit areas (21) each including a plurality of polarizers (22, 22a to 22d) are repeatedly arranged;
An inspection device (30),
The plurality of illumination units (10, 10a to 10o) emit illumination light with different polarization states,
The plurality of polarizers (22, 22a to 22d) transmit light having different polarization directions,
the polarization camera (20) captures an image while the plurality of illumination units (10, 10a to 10o) are simultaneously turned on, thereby outputting a plurality of polarization images (50, 50a to 50d) corresponding to the plurality of polarizers (22, 22a to 22d), respectively;
The inspection system (1) includes an inspection device (30) that inspects the object (W) using the multiple polarized images (50, 50a to 50d).

(Configuration 2)
The plurality of illumination units are arranged such that their azimuth angles around the optical axis (25) of the polarization camera (20) are different from each other;
The plurality of illumination units include first to Nth illumination units (10a to 10d),
The plurality of polarizers include first to Nth polarizers (22a to 22d),
N is an integer of 2 or more,
the polarization directions of the illumination light from the first to Nth illumination units (10a to 10d) are parallel to the polarization directions of the light transmitted through the first to Nth polarizers (22a to 22d), respectively;
The inspection system (1) described in configuration 1, wherein the inspection device (30) generates normal images (51, 52) indicating the normal direction of the surface of the object (W) from the multiple polarized images (50a to 50d) and inspects the object (W) based on the normal images (51, 52).

(Configuration 3)
N is 4,
the first illumination unit (10a) and the third illumination unit (10c) are disposed at positions symmetrical with respect to an optical axis (25) of the polarization camera (20);
the second illumination unit (10b) and the fourth illumination unit (10d) are disposed at positions symmetrical with respect to an optical axis (25) of the polarization camera (20);
a difference between a first azimuth angle at which the first illumination unit (10a) is arranged and a second azimuth angle at which the second illumination unit (10b) is arranged is 90° around an optical axis (25) of the polarization camera (20);
the plurality of polarization images include first to fourth polarization images (50a to 50b) corresponding to the first to fourth polarizers (22a to 22d), respectively;
The inspection device (30)
generating a first normal image (51) indicating a magnitude of a component along the direction of the first azimuth angle in a normal vector of a surface of the object (W) based on the first to fourth polarization images (50a to 50b);
generating a second normal image (52) indicating a magnitude of a component along the second azimuth angle in a normal vector of the surface of the object (W) based on the first to fourth polarization images;
generating a shape image (53) indicating a shape of the surface of the object (W) based on the first normal image (51) and the second normal image (52);
3. The inspection system (1) of configuration 2, for inspecting the object (W) based on the shape image (53).

(Configuration 4)
The inspection system (1) of configuration 3, wherein the angles between the polarization direction of the illumination light of the second illumination section (10b), the third illumination section (10c) and the fourth illumination section (10d) and the polarization direction of the illumination light of the first illumination section (10a) are 45°, 90° and 135°, respectively.

(Configuration 5)
The inspection system (1) according to configuration 1, wherein the plurality of illumination units (10e to 10l) are arranged so as to have different elevation angles relative to the object.

(Configuration 6)
The plurality of illumination units include
A first illumination unit (10e) that irradiates illumination light along an optical axis of the polarization camera;
a ring-shaped second illumination unit (10f) centered on the optical axis of the polarization camera;
The plurality of polarizers includes a first polarizer (22a) and a second polarizer (22c);
The inspection system (1) of configuration 5, wherein the polarization direction of the illumination light emitted from the first illumination section (10e) and the second illumination section (10f) coincides with the polarization direction of the light transmitted through the first polarizer (22a) and the second polarizer (22c), respectively.

(Configuration 7)
The plurality of illumination units include a plurality of ring-shaped illumination units (10g to 10l) concentrically arranged around the optical axis of the polarization camera,
The inspection device (30)
generating a phase image indicating an angle between a normal direction of a surface of the object (W) and an optical axis direction of the polarization camera based on the plurality of polarization images (50a to 50d);
6. The inspection system (1) of configuration 5, for inspecting the object (W) based on the phase image.

(Configuration 8)
The inspection system (1) of configuration 7, wherein the plurality of illumination units further includes an illumination unit (10e) that irradiates illumination light along an optical axis (25) of the polarization camera (20).

(Configuration 9)
The plurality of illumination units include
A first illumination unit (10e) that irradiates illumination light along an optical axis of the polarization camera;
A second illumination unit (10m) that irradiates illumination light from a back side of the object,
The plurality of polarizers includes a first polarizer (22a) and a second polarizer (22c);
The inspection system (1) of configuration 1, wherein the polarization direction of the illumination light emitted from the first illumination section (10e) and the second illumination section (10m) coincides with the polarization direction of the light transmitted through the first polarizer (22a) and the second polarizer (22c), respectively.

(Configuration 10)
The plurality of illumination units include a first illumination unit (10n) that irradiates linearly polarized illumination light and a second illumination unit (10o) that irradiates unpolarized light,
The plurality of polarizers include
a first polarizer (22a) that transmits light having the same polarization direction as the polarization direction of the illumination light from the first illumination unit (10n);
a second polarizer (22b) that transmits light having a polarization direction that forms an angle of 45° with the polarization direction of the light transmitted through the first polarizer (22a);
a third polarizer (22c) that transmits light having a polarization direction that forms an angle of 90° with the polarization direction of the light transmitted through the first polarizer (22a);
the plurality of polarization images include first to third polarization images (55a to 55c) corresponding to the first to third polarizers (22a to 22c), respectively;
The inspection device (30)
The first to third polarization images (55a to 55c) are subjected to high dynamic range synthesis to generate a synthetic image;
2. The inspection system (1) of configuration 1, for inspecting the object (W) based on the composite image.

(Configuration 11)
An inspection method using a plurality of illumination units (10, 10a to 10o) for illuminating an object (W) and a polarization camera (20) in which unit areas (21) including a plurality of polarizers (22, 22a to 22d) are repeatedly arranged,
The plurality of illumination units (10, 10a to 10o) emit illumination light with different polarization states,
The plurality of polarizers (22, 22a to 22d) transmit light having different polarization directions,
The inspection method includes:
acquiring a plurality of polarized images (50, 50a to 50d) corresponding to the plurality of polarizers (22, 22a to 22d), respectively, by capturing an image of the object (W) using the polarization camera (20) while the plurality of illumination units (10, 10a to 10o) are simultaneously turned on;
and inspecting the object (W) using the plurality of polarized images (50, 50a to 50d).

Although the embodiments of the present invention have been described, the embodiments disclosed herein should be considered to be illustrative and not restrictive in all respects. The scope of the present invention is defined by the claims, and it is intended to include all modifications within the meaning and scope of the claims.

1 Inspection system, 2 Conveyor belt, 10, 10a to 10o Illumination unit, 11a to 11o Light emission unit, 12a to 12n Linear polarizing filter, 13e Half mirror, 20 Polarization camera, 21 Unit area, 22, 22a to 22d Polarizer, 25, 225 Optical axis, 30 Inspection device, 50, 50a to 50d Polarized image, 51 X-direction normal image, 52 Y-direction normal image, 53 Shape image, 54 Binary image, 55 Albedo image, 56 Average image, 57 Composite image, 110 Illumination device, 110a to 110d Arc area, 302 Display unit, 304 Keyboard, 306 Memory card, 310 Processor, 312 RAM, 314 Display controller, 316 System controller, 318 Controller, 320 Hard disk, 322 Camera interface, 324 input interface, 328 communication interface, 330 memory card interface, 350 inspection program, F1, F2 frame line, P point, Q focus point, R1 to R4 rectangular area, W object, n normal vector.

Claims (7)

1. An inspection system comprising:
A plurality of illumination units for illuminating an object;
a polarization camera in which unit areas each including a plurality of polarizers are repeatedly arranged;
An inspection device,
The plurality of illumination units emit illumination light having different polarization states,
the plurality of polarizers transmit light having different polarization directions;
the polarization camera captures images while the plurality of illumination units are simultaneously turned on, thereby outputting a plurality of polarization images respectively corresponding to the plurality of polarizers;
The inspection device inspects the object using the plurality of polarized images;
the plurality of illumination units are arranged such that their azimuth angles around the optical axis of the polarization camera are different from one another;
The plurality of illumination units include first to Nth illumination units,
The plurality of polarizers includes first to Nth polarizers,
N is an integer of 2 or more,
the polarization directions of the illumination light of the first to Nth illumination units are parallel to the polarization directions of the light transmitted through the first to Nth polarizers, respectively;
The inspection device includes:
estimating a normal direction of the surface of the object from the plurality of polarized images using photometric stereo;
generating a normal image indicating the normal direction;
An inspection system that inspects the object based on the normal image . N is 4,
the first illumination unit and the third illumination unit are disposed at positions symmetrical with respect to an optical axis of the polarization camera,
the second illumination unit and the fourth illumination unit are disposed at positions symmetrical with respect to an optical axis of the polarization camera,
a difference between a first azimuth angle at which the first illumination unit is arranged and a second azimuth angle at which the second illumination unit is arranged is 90° around an optical axis of the polarization camera;
the plurality of polarization images include first to fourth polarization images corresponding to the first to fourth polarizers, respectively;
The inspection device includes:
generating a first normal image indicating a magnitude of a component along a direction of the first azimuth angle in a normal vector of the surface of the object based on the first to fourth polarization images;
generating a second normal image indicating a magnitude of a component along a direction of the second azimuth angle in a normal vector of the surface of the object based on the first to fourth polarization images;
generating a shape image indicating a shape of a surface of the object based on the first normal image and the second normal image;
The inspection system of claim 1 , further comprising: a sensor configured to detect the object based on the shape image. 3. The inspection system of claim 2, wherein angles between the polarization direction of the illumination light of the second illumination section, the third illumination section, and the fourth illumination section and the polarization direction of the illumination light of the first illumination section are 45°, 90°, and 135°, respectively. 1. An inspection system comprising:
A plurality of illumination units for illuminating an object;
a polarization camera in which unit areas each including a plurality of polarizers are repeatedly arranged;
An inspection device,
The plurality of illumination units emit illumination light having different polarization states,
the plurality of polarizers transmit light having different polarization directions;
the polarization camera captures images while the plurality of illumination units are simultaneously turned on, thereby outputting a plurality of polarization images respectively corresponding to the plurality of polarizers;
The inspection device inspects the object using the plurality of polarized images;
The plurality of illumination units are arranged so as to have different elevation angles with respect to the object ,
The plurality of illumination units include
a first illumination unit that irradiates illumination light along an optical axis of the polarization camera;
a ring-shaped second illumination unit centered on the optical axis of the polarization camera;
the plurality of polarizers includes a first polarizer and a second polarizer;
a polarization direction of illumination light emitted from the first illumination unit and a polarization direction of light transmitted through the first polarizer and the second polarizer , respectively;
The inspection device includes:
inspecting a defect that is less likely to be specularly reflected than a surrounding normal portion based on a polarized image corresponding to the first polarizer among the plurality of polarized images;
An inspection system that inspects for defects that are more likely to reflect diffusely than surrounding normal portions based on a polarized image corresponding to the second polarizer among the plurality of polarized images . 1. An inspection system comprising:
A plurality of illumination units for illuminating an object;
a polarization camera in which unit areas each including a plurality of polarizers are repeatedly arranged;
An inspection device,
The plurality of illumination units emit illumination light having different polarization states,
the plurality of polarizers transmit light having different polarization directions;
the polarization camera captures images while the plurality of illumination units are simultaneously turned on, thereby outputting a plurality of polarization images respectively corresponding to the plurality of polarizers;
The inspection device inspects the object using the plurality of polarized images;
The plurality of illumination units are arranged so as to have different elevation angles with respect to the object ,
the plurality of illumination units include a plurality of ring-shaped illumination units concentrically arranged around an optical axis of the polarization camera;
The inspection device includes:
The plurality of polarized images are regarded as a plurality of images in which stripe patterns having different phases from each other are captured,
generating a phase image indicating an angle between a normal direction of the surface of the object and an optical axis direction of the polarization camera based on the plurality of polarization images using a phase shift method ;
An inspection system that inspects the object based on the phase image. The inspection system according to claim 5, wherein the plurality of illumination units further includes an illumination unit that irradiates illumination light along the optical axis of the polarization camera. An inspection method using a plurality of illumination units that illuminate an object and a polarization camera in which unit areas each including a plurality of polarizers are repeatedly arranged, comprising:
The plurality of illumination units emit illumination light having different polarization states,
the plurality of polarizers transmit light having different polarization directions;
The inspection method includes:
acquiring a plurality of polarization images corresponding to the plurality of polarizers by capturing an image of the object using the polarization camera while the plurality of illumination units are simultaneously turned on;
and inspecting the object using the plurality of polarized images.
the plurality of illumination units are arranged such that their azimuth angles around the optical axis of the polarization camera are different from one another;
The plurality of illumination units include first to Nth illumination units,
The plurality of polarizers includes first to Nth polarizers,
N is an integer of 2 or more,
the polarization directions of the illumination light of the first to Nth illumination units are parallel to the polarization directions of the light transmitted through the first to Nth polarizers, respectively;
The step of inspecting includes:
estimating a surface normal direction of the object from the plurality of polarized images using photometric stereo; and
generating a normal image indicating the normal direction;
and inspecting the object based on the normal image .

Images