JP6944355B2 - Railroad vehicle image generator and railroad vehicle visual inspection system - Google Patents

Description

The present invention relates to an image generator for generating an image of a railroad vehicle using a line scan camera and an appearance inspection system for the railroad vehicle.

Conventionally, it has been proposed to inspect the appearance of a railway vehicle using a photographed image of the railway vehicle (see, for example, Patent Documents 1 and 2). The technique of Patent Document 1 is to photograph a specific part of a railway vehicle from a predetermined angle with a dedicated camera, compare the photographed image with a reference image, and determine the presence or absence of an abnormality. The technique of Patent Document 2 is abnormal by taking a picture of a railroad vehicle running on a railroad track, synthesizing an image of the entire vehicle, setting a predetermined inspection area in the image, and comparing it with an image of a normal railroad car. It determines the presence or absence. The image of the entire railroad vehicle is created by cutting and pasting a plurality of images obtained by high-speed continuous shooting from the side of the railroad vehicle.

Japanese Unexamined Patent Publication No. 5-143714 Japanese Unexamined Patent Publication No. 2013-53875

In the configuration of the invention of the visual inspection of Patent Document 1 and Patent Document 2, an area sensor camera is used as a photographing device. The area sensor camera is a photographing device that images an image of a two-dimensional shooting area in one shooting. However, the area sensor camera has a problem that it is difficult to acquire a captured image having a uniform image quality in the entire range for a captured object that is long in one direction. For example, with an area sensor camera, an image of the shooting target viewed from the front can be obtained at the center of the shooting area, but an image of the shooting target viewed at an angle can be obtained at the edge of the shooting area. Makes a difference. For this reason, even if the captured images obtained by taking a plurality of shots from the start to the end of the railroad vehicle are cut and pasted and the continuous images are combined, the appearance of the railroad vehicle is not uniform depending on the location where the images are cut and pasted.

Therefore, the present inventors have studied a system in which a railroad vehicle is photographed by using a line scan camera (also referred to as a "line sensor camera") and the appearance of the railroad vehicle is inspected using the photographed image. A line scan camera is a shooting device having a linear shooting area, and shoots by continuously shooting a moving shooting target and synthesizing the acquired linear image data so as to be stitched together. A two-dimensional photographed image of the target can be obtained. In general, a line scan camera has an advantage that it is easier to acquire a high-resolution image than an area sensor, and it is easy to acquire an image with less distortion due to lens aberration over the entire range for a long object to be photographed in one direction. be.

However, when a railroad vehicle moving on a rail is photographed using a line scan camera, the scale of the synthesized two-dimensional photographed image in the lateral direction (traveling direction of the railroad vehicle) is reduced due to the speed change of the railroad vehicle. It became clear that there was a problem that it was different for each place.

In addition, in the system for visual inspection of railway vehicles under development, we are considering correcting the scale variation of each part of the photographed image by using a reference image without scale variation. Therefore, it is necessary to create a reference image in which the scale variation is removed for each railroad vehicle, but it has not been easy to create such a reference image. For example, an operator compares a photographed image of a line scan camera with an actual photograph, etc., determines the scale variation of each part of the photographed image by relying on the senses, and stretches and adjusts each part of the photographed image. It required complicated and time-consuming work. Therefore, there is a problem that the accuracy of the generated reference image varies depending on the skill of the operator.

The present invention provides an image generation device for a railway vehicle, which can easily generate a photographed image of the railway vehicle in which the scale of each part is made uniform with high accuracy by photographing the railway vehicle using a line scan camera. The purpose is to do. Furthermore, it is an object of the present invention to provide a railroad vehicle visual inspection system capable of efficiently and highly accurate visual inspection of a railroad vehicle using this photographed image as a reference image.

In order to achieve the above object, the present invention
An image generator for a railroad vehicle that generates a photographed image of a railroad vehicle whose side surface is included in the visual inspection.
A line scan camera in which the linear shooting area intersects the moving direction of the railroad vehicle,
A design data storage unit that stores at least a part of the design data of the railway vehicle, and
A plurality of structures provided in the railway vehicle are specified as a plurality of landmark points on a two-dimensional photographed image of the railway vehicle synthesized by connecting image data obtained by continuous photographing by the line scan camera. Mark point identification processing unit and
Based on the plurality of landmark points in the identified photographed image and the arrangement information of the plurality of structures in the design data, the scale of the railroad vehicle in the traveling direction in the photographed image is made uniform. An image correction unit that expands and contracts each part of the captured image,
It is an image generation device of a railroad vehicle characterized by being provided with.

According to this configuration, by photographing a railway vehicle with a line scan camera, a two-dimensional photographed image of a high-resolution railway vehicle with less distortion due to lens aberration can be acquired as compared with the case of using an area sensor. Can be done. Further, the image correction unit measures the scale of the traveling direction of the railway vehicle in the photographed image based on the arrangement information of the plurality of structures shown in the design data of the railway vehicle and the plurality of landmark points identified from the photographed image. Stretch each part of the captured image so that it is uniform. As a result, the lateral scale variation of the captured image due to the speed change of the railroad vehicle is corrected, and the scale of each part is high in a short time without relying on human skill and without complicated work. It is possible to generate a photographed image of a railroad vehicle that has been made uniform with accuracy.

Here, the line scan camera is installed in so that contains the vehicle body bottom and sides of the railway vehicle in the photographed image,
The mark point identification processing unit applies a plurality of beams located on the bottom surface of the vehicle body as the plurality of structural parts.
The image correction unit may adopt a configuration in which the photographed image is expanded and contracted including the side surface of the railway vehicle based on the plurality of mark points and the arrangement information of the plurality of beams located on the bottom surface of the vehicle body. good.

Generally, a plurality of accessories that require visual inspection are provided on the lower part of the vehicle body, while a plurality of beams are stretched on the bottom surface of the vehicle body. Therefore, in the above configuration, the line scan camera is installed so as to photograph the lower part of the vehicle body of the railway vehicle obliquely upward from the side. As a result, the accessory provided at the lower part of the vehicle body and the plurality of beams stretched over the bottom surface of the vehicle body are reflected in the photographed image. Then, the image correction unit performs image correction using the positions of the plurality of beams as mark points, so that the scale of each part is uniformly corrected with high accuracy for the captured image in which the plurality of accessories at the lower part of the vehicle body are captured. be able to.

Further, in the image generation device for a railroad vehicle according to the present invention, the image correction unit expands and contracts the captured image including an image region having a height different from the portion of the captured image in which the plurality of beams are reflected. It may be configured to do so.
According to this configuration, it is possible to correct the scale of the side surface of the railroad vehicle with high accuracy based on the arrangement information of a plurality of beams at the lower part of the car body.

Further, the image generation device for a railway vehicle according to the present invention may further include a mark point setting processing unit capable of setting the plurality of mark points by an operation input from the outside.
According to this configuration, when there is an appropriate alignment structure according to the railway vehicle, such as the end of the bogie of the railway vehicle or the air-conditioned housing on the vehicle ceiling, this is performed via the mark point setting processing unit. Can be set as a landmark point. Then, the scale of the captured image can be corrected with high accuracy by using these mark points.

In addition, the visual inspection system for railway vehicles according to the present invention is
The above-mentioned image generation device for a railway vehicle, which generates the photographed image uniformly corrected in scale by the image correction unit as a reference image for the railway vehicle.
The inspection target captured by the line scan camera is compared with the reference image, and the scale of the railway vehicle in the traveling direction is corrected with respect to the captured image of the inspection target, and the corrected inspection target is corrected. An appearance inspection device that inspects the appearance of the railway vehicle using the photographed image of
It is characterized by having.
According to this configuration, the scale of the photographed image to be inspected can be corrected by using the photographed image generated by the image generator as a reference image, and the appearance inspection of the railway vehicle can be performed with high accuracy.

According to the railway vehicle image generation device according to the present invention, the railway vehicle is photographed by using the line scan camera, and the photographed image of the railway vehicle in which the scale of each part in the image is made uniform with high accuracy can be easily generated. can do. Further, according to the railway vehicle appearance inspection system according to the present invention, the appearance inspection of a railway vehicle can be efficiently performed with high accuracy by using a photographed image in which the scale of each part in the image is made uniform with high accuracy as a reference image. It can be carried out.

It is a block diagram which shows the appearance inspection system of the railroad vehicle which concerns on embodiment of this invention. It is a figure which shows an example of the photographed image acquired by the line scan camera. It is a flowchart which shows the procedure of the reference image generation processing executed by the image processing apparatus for reference image generation. It is a figure explaining the reference image generation process of FIG. It is a flowchart which shows the procedure of the mark point setting process executed by the image processing apparatus for the reference image generation. It is a flowchart which shows the procedure of the appearance inspection processing executed by the image processing apparatus for appearance inspection. It is a flowchart which shows the procedure of the scale correction processing executed in step S43 of FIG. It is a figure explaining the scale correction process of FIG.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.
FIG. 1 is a configuration diagram showing an appearance inspection system for a railway vehicle according to an embodiment of the present invention.
The railroad vehicle visual inspection system according to the embodiment of the present invention includes a line scan camera 3, an image processing device 1 for generating a reference image, and an image processing device 2 for visual inspection. Of these, the combination of the line scan camera 3 and the image processing device 1 corresponds to an example of the image generation device for a railway vehicle according to the present invention. Further, the combination of the line scan camera 3 and the image processing device 2 corresponds to an example of the visual inspection device for a railway vehicle according to the present invention.

The line scan camera 3 is an imaging device that acquires image data of a linear imaging area in one imaging. The line scan camera 3 is installed so that the linear photographing area intersects the moving direction of the railway vehicle 300, and continuously and repeatedly performs photographing while the railway vehicle 300 is moving. By synthesizing the image data acquired by continuous photographing, a two-dimensional photographed image of the railway vehicle 300 can be obtained.

The image processing device 1 for generating a reference image is a computer, and is a CPU (Central Processing Unit) 10, a RAM (Random Access Memory) 11, a display unit 12, an operation unit 13 such as a mouse or keyboard, and an interface 14 for data input. , And a storage device 15. The storage device 15 is set with a reference image generation program 151 that generates a reference image corrected in the lateral scale from the captured image acquired by the line scan camera 3, and a mark point used for generating the reference image. The mark point setting program 152 to be used is stored. Further, the storage device 15 is photographed by a design data storage unit 153 that stores a part of the design data of the railroad vehicle 300, a reference image storage unit 154 that stores the generated reference image, and a line scan camera 3. A captured image storage unit 155 for storing the image data acquired in the above is provided. The captured image storage unit 155 stores image data sent from the line scan camera 3 via, for example, the communication network N. In the above configuration, the CPU 10 that executes the reference image generation program 151 functions as a mark point identification processing unit and an image correction unit according to the present invention. Further, the CPU 10 that executes the mark point setting program 152 functions as a mark point setting processing unit according to the present invention.

The image processing device 2 for visual inspection is a computer, and includes a CPU 20, a RAM 21, a display unit 22, an operation unit 23 such as a mouse or a keyboard, an interface 24 for data input, and a storage device 25. The storage device 25 stores a visual inspection program 251 that performs a visual inspection of the railway vehicle 300 based on a photographed image for visual inspection acquired by the line scan camera 3. Further, the storage device 25 is subjected to an appearance inspection, a photographed image storage unit 252 for storing image data acquired by the line scan camera 3 and a reference image storage unit 253 for storing a reference image of each railroad vehicle 300. It includes an image database 254 that manages images of the railcars 300 that have passed through. The reference image storage unit 253 stores in advance the image data of the reference image generated by the image processing device 1 for generating the reference image. The image data of the reference image may be sent from the image processing device 1 for generating the reference image to the image processing device 2 for visual inspection via the communication network N or the storage medium. In the image database 254, the photographed image of the railroad vehicle 300 subject to the visual inspection is stored in association with, for example, the identification information of the railroad vehicle 300 and the inspection result information.

<Shooting process with line scan camera>
FIG. 2 is a diagram showing an example of a captured image acquired by a line scan camera.
The line scan camera 3 is installed on the side of the rail through which the railroad vehicle 300 passes, facing diagonally upward. Combining the image data obtained by the continuous shooting of the line scan camera 3 when the railroad vehicle 300 is moving, as shown in FIG. 2, the shot image including the region where the lower part of the vehicle body of the railroad vehicle 300 is viewed diagonally upward. F0 is obtained. A plurality of beams 302 stretched in the horizontal direction (direction orthogonal to the traveling direction of the railway vehicle 300) are reflected on the bottom surface of the vehicle body in the captured image F0. Although FIG. 2 mainly shows the lower portion of the vehicle body in the captured image F0, the captured image F0 may include the entire side surface of the vehicle body (see FIG. 8A).

If the moving speed of the railroad vehicle 300 varies during the shooting of the line scan camera 3, the scale of the captured image in the lateral direction (traveling direction of the railroad vehicle 300) varies. Further, during the shooting of the line scan camera 3, if the vertical displacement of the rail due to the passage of the railroad vehicle 300 and the vertical movement of the vehicle body due to the action of the suspension such as the air spring occur, the captured image of the railroad vehicle 300 is distorted in the vertical direction. Occurs.

<Reference image generation process>
Subsequently, the reference image generation process executed by the image processing device 1 for generating the reference image will be described. The reference image generation process is realized by the CPU 10 executing the reference image generation program 151 of FIG. FIG. 3 is a flowchart showing a procedure of a reference image generation process executed by the image generation device of the embodiment. FIG. 4 is an explanatory diagram showing the flow of the reference image generation process.

In order to start the reference image generation process, as a preparation, first, a railroad car 300 connected to one train in a state where there is no abnormality in appearance is run on the rail, and the railroad car 300 is photographed by the line scan camera 3. conduct. Then, the operator or the like stores the captured image in the captured image storage unit 155 of the image processing device 1. Further, the design data storage unit 153 stores design data showing arrangement information of a plurality of beams 302 under the vehicle body of the railway vehicle 300. In this state, for example, when the operator inputs a start command to the image processing device 1 via the operation unit 13, the reference image generation process is started.

When the reference image generation process is started, the CPU 10 first reads the image data of the photographed image of the railway vehicle 300 from the photographed image storage unit 155 (step S1), and the railway vehicles 300 of a plurality of vehicles connected to one train set. Images are combined (step S2). The image data obtained from the line scan camera 3 is data in which a plurality of sets of image data of one vertical line are collected, and a two-dimensional captured image is combined by synthesizing these data.
Subsequently, the CPU 10 performs brightness correction and ground shaking correction on the captured image (step S3). In the captured image, the brightness and contrast of the entire image vary depending on the amount of external light at the time of photographing. Further, the height of the rail of the bogie portion is slightly lowered due to the weight of the railroad vehicle 300, so that the captured image is distorted in the vertical direction. The brightness correction in step S2 removes the variation in brightness and contrast, and the ground shaking correction removes the vertical distortion based on the displacement of the rail. The ground shaking correction is, for example, an image process in which the captured image is vertically displaced for each vertical line so that the rails on the captured image are linear.

After performing the correction processing, the CPU 10 cuts out the captured images for one train set into the captured images for each of the railway vehicles 300 (step S4), and selects the captured images for one of the trains as the processing target (step S5). ).
When the captured image to be processed is selected, the CPU 10 performs a process of specifying a plurality of preset mark points from the captured image (step S6). The process of step S6 corresponds to an example of the process by the mark point identification processing unit according to the present invention. In the present embodiment, a plurality of beams 302 stretched over the lower part of the vehicle body are set as mark points. In the mark point identification process, for example, the CPU 10 outputs a captured image to the display unit 12, outputs a message so that the operator points to a plurality of beams 302, and an operation input for pointing the beam 302 from the operator via the operation unit 13. It can be realized by the process of accepting. In this case, the CPU 10 recognizes the point pointed out by the operation input as a mark point. Alternatively, the mark point identification process may be realized by the CPU 10 performing an image recognition process using the local image feature amount of the beam 302 stored in advance. Alternatively, the image recognition and the input processing of the operator may be combined so that the CPU 10 specifies the position of the beam 302.

Subsequently, the CPU 10 divides the captured image to be processed into, for example, a rectangular frame having a total number of pixels in the vertical direction and a predetermined number of images in the horizontal direction, and cuts out a plurality of strip images (step S7). The upper part P1 of FIG. 4 shows an example of some cut-out strip images F1. The number of pixels on the side of the strip image F1 may be a fixed number, or may be adjusted to a different number of pixels for each strip image F1 so that the mark points are included in each strip image F1.

After cutting out the strip image F1, the CPU 10 acquires the arrangement information of the plurality of corresponding beams 302 from the design data stored in the design data storage unit 153 (step S8). The middle part P2 of FIG. 4 shows a diagram conceptually representing the design data. In step S8, the CPU 10 may calculate the position of the beam 302 from the dimensional data of each part of the design data and use this as the arrangement information. Alternatively, if the design data includes the data of the design drawing, the CPU 10 may recognize the position of the beam 302j from the design drawing by processing such as image recognition, and use this recognition location as the arrangement information of the beam 302j. Instead of image recognition, the CPU 10 outputs a design drawing to the display unit 12, outputs a message to the operator so as to point to the beam 302j, and receives an operation input indicating the position from the operator via the operation unit 13. You may go. Even by such processing, the CPU 10 can acquire the arrangement information of the plurality of beams 302j on the design drawing.

After acquiring the arrangement information, the CPU 10 expands and contracts the width of each strip image F1 in the horizontal direction so as to match the mark points on the captured image with the design arrangement (step S9). The process of step S9 corresponds to an example of the process of the image correction unit according to the present invention. A plurality of strip images F2 after the expansion / contraction processing are shown in the middle portion P3 of FIG. The plurality of strip images F2 are expanded and contracted so as to align each mark point of the strip image F1 shown in the upper portion P1 with the arrangement of the beam 302j on the design data. The original strip image F1 includes a variation in the lateral scale according to the speed variation of the railway vehicle 300 at the time of shooting. On the other hand, the strip image F2 after expansion and contraction is expanded and contracted from the width dimensions L1 to L6 of the original strip image F1 to the width dimensions L1a to L6a of the strip image F2 so that the variation in scale is removed.

When the expansion and contraction of each strip image F2 is completed, the CPU 10 connects all the strip images F2 for one car and synthesizes the photographed image F3 for the railroad vehicle 300 for one car. The lower part P4 of FIG. 4 shows an example of the captured image F3 after composition. The photographed image of FIG. 4 mainly shows the lower part of the vehicle body of the railway vehicle 300, but the photographed image and the strip image may include an image of the entire side surface of the vehicle body (see FIG. 8A). In that case, by the above processing, a photographed image having a uniform lateral scale can be obtained with respect to the photographed image extending from the lower part of the vehicle body to the entire side surface of the railway vehicle 300.

When the captured images F3 of the railway vehicle 300 for one car are combined, the CPU 10 determines whether or not the processing for the captured images for all the vehicles in one train has been completed (step S11), and there is a captured image of the unprocessed vehicle. If so, the process returns to step S5. Then, the CPU 10 executes the processes of steps S5 to S10 for the captured image of the unprocessed vehicle.
On the other hand, when the processes of steps S4 to S10 are completed for the captured images of all the vehicles in one organization and YES is determined in the determination process of step S11, the CPU 10 connects the captured images F3 of all the vehicles (step). S12). As a result, a reference image of one train of railway vehicles 300 is generated, and the reference image generation process is completed.

<Mark point setting process>
In the above reference image generation process, an example of generating a reference image using a plurality of beams 302 at the lower part of the vehicle body as mark points is shown. However, as the mark point, a new mark point may be set by input from the outside by the mark point setting process described later.
FIG. 5 is a flowchart showing a procedure of a mark point setting process executed by an image processing device for generating a reference image. The mark point setting process corresponds to an example of the process of the mark point setting processing unit according to the present invention, and is realized by executing the mark point setting program 152 of FIG. 1 by the CPU 10. In the following, as an example, in the reference image generation process of FIG. 3, the mark points are automatically specified by image recognition, and the mark point arrangement information is automatically calculated from the design data. The mark point setting process that can be performed will be described.

The mark point setting process is started when the operator inputs a start command via the operation unit 13 of the image processing device 1. When the mark point setting process is started, the CPU 10 first displays a photographed image of the railway vehicle 300 on the display unit 12 (step S21), and then the display image is displayed by the operator's input via the operation unit 13. Performs an input process for pointing to a new mark point (step S22). Here, the operator uses the mouse of the operation unit 13 to perform an operation input pointing to a portion (for example, the left end or the right end of the bogie) that can be a new mark point in the photographed image of the railway vehicle 300.

After performing the input process, the CPU 10 subsequently calculates the new mark point pointed to and the local image feature amount around it, and saves the calculated local image feature amount and the location section of the new mark point. (Step S23). Here, as the local image feature amount, for example, HOG (Histograms of Oriented Gradients) used for image pattern recognition or the like can be applied. The location section of the mark point represents any one of the captured images for one car, for example, roughly divided into about 3 × 3.

Next, the CPU 10 performs image recognition of a new mark point from the saved local image feature amount and the information of the location section, and displays the mark point of the recognition result in the captured image of the display unit 12, for example, by displaying the primary colors. Output (step S24). Then, the CPU 10 inquires the operator whether or not the recognition result is correct, and waits for a reply from the operator (step S25). Here, the operator inputs OK if the mark points by image recognition match those input in step S22, and NG if they do not match or if a plurality of mark points are recognized. Enter. As a result, if it is NG, the CPU 10 displays on the display unit 12 that the setting is not possible (step S31), and ends the mark point setting process. On the other hand, if it is OK, the CPU 10 proceeds to the next step.

Next, when the processing proceeds, the CPU 10 displays an image of the design drawing representing the design data on the display unit 12 (step S26), and subsequently, by inputting by the operator via the operation unit 13, the display image is displayed. An input process for pointing to the location of the structure corresponding to the mark point is performed (step S27). Here, the operator uses the mouse of the operation unit 13 to perform an operation input pointing to a corresponding mark point in the design drawing.

After performing the input processing, the CPU 10 subsequently calculates the width length (width length along the traveling direction of the railway vehicle 300) from the reference point (for example, the head of the vehicle body) of the railway vehicle 300 to the pointed point from the design data. The calculation is performed, and the calculation result is output to the display unit 12 (step S28). Then, the CPU 10 inquires the operator whether or not the calculation result is correct, and waits for a reply from the operator (step S29). Here, the operator calculates the width length from the reference point to the mark point of the railway vehicle 300 using the value shown in the design data, collates it with the displayed calculation result, and inputs OK if they match. If there is a discrepancy, input NG.

As a result, if it is NG, the CPU 10 displays on the display unit 12 that the setting is not possible (step S31), and ends the mark point setting process. On the other hand, if it is OK, the CPU 10 uses the information of the local image feature amount and the location section saved in step S23 and the calculation function of step S28 (which value in the design data is used to calculate with what calculation formula). (Function indicating) is stored as setting information of a new mark point (step S30). Then, the mark point setting process is completed.
In the above-mentioned mark point setting process, a processing example is shown in which the CPU 10 obtains a calculation function of the width length from the reference point of the railway vehicle 300 to the new mark point. However, when the design data is complicated, the operator inputs this calculation function (a function indicating which value in the design data is used and what kind of calculation formula is used) to the image processing device 1. You may do so.

By adding the mark points by the mark point setting process, the number of mark points recognized in the image in step S6 of the above-mentioned reference image generation process (FIG. 3) increases, and the scale of each part of the captured image varies in more detail. Can be corrected to remove. Further, when a reference image is generated for a photographed image of a portion where the beam 302 at the lower part of the vehicle body is not reflected, such as on the roof of the railway vehicle 300, the edge of the air-conditioning housing on the ceiling of the vehicle body, the antenna housing, for example, is generated by the mark point setting process. A plurality of landmarks such as the edge of the body may be set. As a result, the reference image generation process described above makes it possible to generate a reference image in which the scale is uniformly corrected even for a photographed image of a portion where the beam 302 at the lower part of the vehicle body is not reflected.

<Appearance inspection processing>
Subsequently, the appearance inspection processing of the railway vehicle 300 executed by the image processing device 2 for appearance inspection using the reference image will be described. There are many variations in the visual inspection method using the reference image, and the following visual inspection processing is only one example. The visual inspection process is realized by the CPU 20 executing the visual inspection program 251 of FIG. FIG. 6 is a flowchart showing a procedure of the visual inspection process executed by the visual inspection apparatus of the embodiment.

In order to start the visual inspection process, as a preparation thereof, a photographed image (photographed image to be inspected) of the railway vehicle 300 taken on the inspection day is stored in the photographed image storage unit 252 and previously generated by the image processing device 1. It is assumed that the reference image of the same railroad vehicle 300 is stored in the reference image storage unit 253. In this state, the visual inspection process is started when the operator inputs a start command via the operation unit 23.

When the appearance inspection process is started, first, the CPU 20 of the image processing device 2 reads out the image data of the photographed image to be inspected from the photographed image storage unit 252 (step S41), corrects the brightness of the photographed image, and shakes the ground. The correction is performed (step S42). The brightness correction and the ground shaking correction are the same processes as those in step S2 (FIG. 3) of the reference image generation process.
Subsequently, the CPU 20 executes the scale correction process on the captured image to be inspected by using the reference image stored in the reference image storage unit 253 (step S43). The details of the scale correction process will be described later.

Next, the CPU 20 corrects the vertical distortion of the captured image due to the vertical movement of the vehicle body with respect to the bogie via a suspension such as an air spring (step S44). Since the vertical distortion of the captured image occurs in the vehicle body portion and not in the trolley portion, for example, the image region of the vehicle body and the accessories fixed to the vehicle body in the captured image is extracted, and the edge of the bottom of the vehicle body is straight. This is achieved by moving the image area extracted so as to be in the vertical direction for each pixel in a row.
By the correction processing of steps S42 to S44, each photographed image is caused by the brightness at the time of shooting, the vertical displacement of the rail due to the passage of the railroad vehicle 300, the speed variation of the railroad vehicle 300, and the vertical movement of the vehicle body with respect to the bogie. The resulting non-uniformity is removed.

Once the photographed image of the railway vehicle 300 corrected in this way is obtained, the CPU 20 next reads out the image data of the past normal image of the same vehicle stored in the normal image database 254A, and photographs this image. Comparison with the image is performed (step S45). The normal image database 254A is provided in the image database 254 of FIG. 1, and is a database in which captured images determined to be normal at a high level in the visual inspection are registered. According to the image comparison in step S45, when there is an abnormality such as a scratch on the side wall of the railway vehicle 300 or the doors of various housings are about to open, such an abnormality is regarded as a difference on the image. Can be detected.

After comparing the images, the CPU 20 determines whether or not there is an abnormal portion in the appearance of the railway vehicle 300 based on the amount of difference on the image (step S46). In the determination process, the CPU 20 may display the determination result on the display unit 22, and when there is an abnormal portion, perform processing such as displaying the abnormal portion in the captured image on the display unit 22.
If it is determined that the abnormality is included as a result of the determination in step S46, the CPU 20 registers the captured image to be inspected in the abnormality image database 254B (step S47). The abnormal image database 254B is provided in the image database 254 of FIG. 1, and the photographed image including the abnormality and the information of the abnormal portion are registered in association with each other.

On the other hand, as a result of the determination in step S46, if it is determined that the appearance of the railway vehicle 300 does not include an abnormal portion, the CPU 20 determines whether or not the conditions for registration in the normal image database 254A are satisfied (step S48). The registration condition is a condition that the degree of matching between the captured image and the normal image is obtained at a high level in the image comparison in step S45. As a result, if the registration condition is not satisfied, the appearance inspection process is terminated as it is, but if the registration condition is satisfied, the CPU 20 registers the captured image in the normal image database 254A (step S49) and performs the appearance inspection. End the process.

<Scale correction processing>
Next, the details of the scale correction process performed in step S43 of the above-mentioned visual inspection process will be described. FIG. 7 is a flowchart showing the procedure of the speed expansion / contraction correction processing. FIG. 8 is an explanatory diagram of the speed expansion / contraction correction process.
When the process shifts to the scale correction process, the CPU 20 first reads the image data of the reference image of the railway vehicle 300 from the reference image storage unit 253 (step S51). Then, a predetermined feature point is extracted from the reference image and the captured image to be inspected (step S52). FIG. 8A shows a reference image F10 in which a predetermined feature point is extracted and a plurality of extracted feature points are connected by a line segment Lr. FIG. 8B shows a captured image F11 to be inspected in which a predetermined feature point is extracted and a plurality of extracted feature points are connected by a line segment Lt. As the feature points, arbitrary points that are easy to identify on the image, such as the edges of each component of the railway vehicle 300, may be set in advance.

Next, as shown in FIG. 8B, the CPU 20 divides the captured image to be inspected into a plurality of rectangular regions f (step S53). The plurality of rectangular regions f are provided so as to divide the width of the photographed image in the lateral direction (traveling direction of the railway vehicle 300) into, for example, a predetermined number of pixel widths.
Subsequently, the CPU 20 calculates the difference between the scale of each rectangular region f and the scale of the corresponding portion of the reference image from the comparison of the feature points (step S54). As a specific example, the CPU 20 first has a line segment Lr connecting a plurality of feature points extracted in step S52 with respect to the captured image to be inspected, and a plurality of feature points extracted in step S52 with respect to the reference image. Create a line segment Lt connecting the above on the data. Then, the CPU 20 compares the slope of the line segment Lt of each rectangular region f with the slope of the line segment Lr of the corresponding region of the reference image. If the horizontal scale of the rectangular area f and the horizontal scale of the reference image are equal, the slopes of the line segments Lt and Lr of the corresponding areas are equal. However, if the lateral scale of the rectangular region f is different to be larger or smaller than the lateral scale of the reference image, the slope of the line segment Lt is larger than the slope of the line segment Lt of the corresponding region of the reference image. Different values will be applied to the smaller one or the larger one. Further, the difference in the slopes of these line segments Lt and Lr depends on the difference in scale between the two regions. Therefore, the CPU 20 determines the horizontal scale of the reference image and the horizontal scale of each rectangular area f from the difference between the slope of the line segment Lt of each rectangular area f and the slope of the line segment Lr of the corresponding area of the reference image. The difference with can be obtained.

After obtaining the difference in scale from the reference image, the CPU 20 expands and contracts each rectangular region f of the captured image to be inspected in the lateral direction at a ratio corresponding to the difference in scale (step S55). The expansion / contraction ratio may be, for example, a ratio such that the scale of the reference image and the scale of the captured image are equal. When the expansion / contraction processing is completed, the CPU 20 synthesizes the images of the plurality of rectangular areas f after expansion / contraction so as to be concatenated (step S56), and ends the scale correction processing. FIG. 8C shows a captured image F12 to be inspected in which images of a plurality of rectangular regions f are expanded and contracted and then connected. By the above processing, the lateral scale variation of the photographed image due to the speed variation of the railway vehicle 300 at the time of photographing can be removed by using the reference image.

As described above, according to the railway vehicle appearance inspection device of the present embodiment, since the railway vehicle is photographed by using the line scan camera 3, the distortion due to the lens aberration is increased as compared with the case where the area sensor is used. It is possible to acquire a small number of high-resolution captured images. Further, in the reference image generation process, the image processing device 1 uses the captured image based on the plurality of beams 302 set as the mark points on the captured image and the arrangement information of the plurality of beams 302 shown in the design data. The captured image is stretched and corrected so as to make the horizontal scale in the above uniform. As a result, the reference image in which the lateral scale variation of the photographed image due to the speed variation of the railway vehicle 300 at the time of shooting is accurately removed can be obtained without relying on the skill of the operator and without complicated work. Can be generated in a short time.

Further, according to the image processing device 1 for generating the reference image of the embodiment, the mark point can be set at a portion other than the beam 302 at the lower part of the vehicle body in the captured image by the mark point setting process. As a result, for example, a mark point can be set in the vicinity of the trolley where the beam 302 is not reflected so much, and the lateral scale variation of the captured image can be removed with high accuracy. Alternatively, even for a photographed image in a region where the beam 302 at the lower part of the vehicle body is not reflected at all, such as a photographed image on the roof of the vehicle body, the lateral scale variation can be removed with high accuracy by the same processing.

Further, according to the visual inspection device for a railway vehicle of the present embodiment, a correction process for removing lateral scale variation of a photographed image to be inspected is performed using the reference image generated as described above. .. Then, the appearance inspection of the railway vehicle 300 is performed by comparing the photographed image after the correction with the normal photographed image of the railway vehicle 300. Therefore, it is possible to accurately remove the lateral scale variation appearing in the photographed image due to the speed variation of the railway vehicle 300 at the time of shooting, and to perform the appearance inspection of the railway vehicle 300 with high efficiency and high accuracy. ..

The embodiment of the present invention has been described above. However, the present invention is not limited to the above embodiment. For example, in the reference image generation process of the above embodiment, as a process by the image correction unit according to the present invention, the captured image is divided into a predetermined width, a plurality of strip images F1 are cut out, and the plurality of strip images F1 are combined. An example of expanding and contracting using a mark point is shown. However, as the processing of the image correction unit according to the present invention, for example, the captured image is divided at the mark points, a plurality of strip images are cut out, and the intervals between the plurality of mark points calculated from the design data are obtained. A process of expanding and contracting according to the length may be applied.

Further, in the above embodiment, an example is shown in which a photographed image having a uniform horizontal scale is generated as a reference image by using the design data, but the photographed image to be inspected is laterally used by using the design data. A process of making the scale of the direction uniform may be applied. In other words, the image generation device for a railway vehicle according to the present invention may be a device for generating a reference image or a device for generating a photographed image to be inspected.

Further, in the above embodiment, an example in which the image processing device 1 for generating the reference image of the railroad vehicle and the image processing device 2 for performing the visual inspection of the railroad vehicle are configured by separate computers is shown, but the reference image generation processing is shown. And the visual inspection process may be performed by one computer. Further, in the above embodiment, an example in which the line scan camera 3 for photographing the railway vehicle 300 for generating a reference image and the line scan camera 3 for photographing the railway vehicle 300 for visual inspection are shown in common. However, these line scan cameras may be provided separately as long as they have common specifications.

Further, the specific procedure of the visual inspection process shown in the above embodiment is only an example, and if the lateral scale correction of the captured image is performed using the reference image, the visual inspection is performed by any procedure. You may. For example, an image comparison may be performed using a reference image as a normal image, and an appearance inspection may be performed. In addition, the details shown in the embodiment can be appropriately changed without departing from the spirit of the invention.

1 Image processing device for reference image generation 2 Image processing device for visual inspection 3 Line scan cameras 10, 20 CPU
11, 21 RAM
12, 22 Display unit 13, 23 Operation unit 14, 24 Interface 15, 25 Storage device 151 Reference image generation program 152 Mark point setting program 153 Design data storage unit 154 Reference image storage unit 155 Photographed image storage unit 251 Visual inspection program 252 Photographing Image storage 253 Reference image storage 254 Image database 300 Railroad vehicle 302 Beam 302j Beam on design data

Claims (2)

An image generator for a railroad vehicle that generates a photographed image of a railroad vehicle whose side surface is included in the visual inspection.
A line scan camera linear imaging region is disposed in a direction intersecting the moving direction of the railway vehicle,
A design data storage unit that stores at least a part of the design data of the railway vehicle, and
A plurality of structures provided in the railway vehicle are specified as a plurality of landmark points on a two-dimensional photographed image of the railway vehicle synthesized by connecting image data obtained by continuous photographing by the line scan camera. Mark point identification processing unit and
Based on the plurality of landmark points in the identified photographed image and the arrangement information of the plurality of structures in the design data, the scale of the railroad vehicle in the traveling direction in the photographed image is made uniform. An image correction unit that expands and contracts each part of the captured image,
Equipped with a,
The line scan camera is installed so that the captured image includes the bottom surface and the side surface of the railroad vehicle.
The mark point identification processing unit applies a plurality of beams located on the bottom surface of the vehicle body as the plurality of structural parts.
The image correcting unit, based on the arrangement information of the plurality of beams positioned on the underbody and the plurality of landmark points, characterized that you stretch the captured image, including the sides of the rail vehicle Image generator for railroad vehicles. The captured image is uniformly corrected scale by the image correction unit, and an image generation apparatus of a railway vehicle according to claim 1 for generating a reference image of the railway vehicle,
The inspection target captured by the line scan camera is compared with the reference image, and the scale of the railway vehicle in the traveling direction is corrected with respect to the captured image of the inspection target, and the corrected inspection target is corrected. An appearance inspection device that inspects the appearance of the railway vehicle using the photographed image of
A railroad vehicle visual inspection system characterized by being equipped with.

Images