JPS6314281B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a pattern inspection method.

When a trademark or the like is printed on the surface of an object to be inspected, such as a crown used in a sake bottle, different kinds of crowns may be mixed in during the manufacturing process, or printing errors or misalignment may occur. Therefore, in this crown manufacturing line, there is a demand for an inspection device that can detect abnormalities such as printing errors, printing misalignment, and mixing of different types of crowns. The present invention provides such two
The present invention relates to a pattern inspection method for identifying patterns that spread across dimensions or detecting normal and abnormal defects.

The problem with constructing an inspection device for the pattern printed on the crown is that when such a circular object to be inspected is flowing on the production line, the direction of the pattern is not the same as shown in Figure 1. As shown in , it is at a point where it becomes indeterminate with respect to rotation.

Generally, two-dimensional pattern inspection equipment uses an industrial camera (ITV) or a similar two-dimensional solid-state image sensor as a sensor, and detects the pixel (i, j) address of a matrix image obtained by these sensors. Pattern inspection is performed by applying signal processing to density information, geometric phase information, and the like. The algorithm for pattern inspection, which is the content of this signal processing, varies depending on each target and manufacturer, but the most commonly used method is to calculate the degree of consistency between a previously registered reference pattern and a sampling pattern to be inspected. There is a pattern matching method that identifies sampling patterns or detects defects based on numerical values representing the degree of matching.

Next, a problem with the random orientation of the pattern will be explained. Considering the pattern matching method, if the inspection pattern is allowed to be rotated, to deal with this, a reference pattern is also prepared that is sequentially rotated by a certain angle, and the sampling pattern and this rotated pattern are prepared. It is necessary to perform inspection by matching all reference patterns that have been created. As a result, the calculation time becomes long, which may make it unsuitable for practical use as an inspection device on a production line, and the adoption of high-performance signal processing equipment to shorten the calculation time may make the inspection device extremely expensive. Something like that happens. Considering that whether a pattern inspection device can be put into practical use is determined by processing time and price, the fact that the inspection pattern has this degree of freedom in rotation is a very important issue in developing pattern inspection devices. It becomes a point.

The present invention relates to the identification or defect inspection of a pattern to be inspected that flows along a line and has a two-dimensional spread that is given a degree of freedom in the rotational direction about the origin. The basic idea of this invention is to define the area distribution, which is an attribute value that does not change with pattern rotation, as an identification parameter. Inspection pattern identification or defect inspection is performed by determining the degree of similarity, that is, the degree of matching, with the area distribution.

As mentioned before, the fact that the inspection pattern has a degree of rotational freedom is a very important factor in configuring the inspection device. Therefore, in this invention,
We defined the area distribution, which is an attribute value that does not change with pattern rotation, and decided to use it as an identification parameter.

In this invention, the area distribution of a pattern means a radius r from the origin as shown in FIGS. 2 and 3.
It is defined as the distribution in the radial direction of the quantity f(r) corresponding to the bright and dark state on the ring band set to ~r+Δr.

It will be understood that this area distribution represents a pattern unique to each pattern, and is a one-dimensional pattern that does not change with respect to rotation of the pattern around the origin. Furthermore, it also has an information compression function that converts a two-dimensional pattern into a one-dimensional pattern called area distribution.

Next, FIG. 4 shows an example of the configuration of the inspection station. An object to be inspected manufactured by the manufacturing apparatus 1 is guided to an inspection station 3 by a conveyance line such as a belt conveyor 2. The inspection station 3 sets an object to be inspected in a setting section 4 and sequentially rotates in a stepwise manner in a step mode. The inspection is performed when the object 6 to be inspected is in the inspection zone 5. This position detection is performed by a rotation pulse of the inspection station or a positioning signal from a separately set photoelectric switch or the like. By configuring the inspection station in this way, (1) it is possible to automatically align the geometric origin of the sensor placed above the inspection zone with the geometric origin of the inspection pattern;

(2) By operating in step mode, the inspection pattern is momentarily stopped. This eliminates movement blur when reading patterns with the sensor.

The inspection station is not limited to that shown in FIG. For example, an inspection station as shown in FIG. 5 may be used. That is, the inspection station 3 is not in step mode but is rotating uniformly and continuously, and the inspection station's rotation pulse or a separately set photoelectric signal indicates that the object to be inspected has arrived at a predetermined position in the inspection zone 5. It is detected using a positioning signal from a switch, etc., and the pattern is read by a sensor. At this time, when the inspection speed is slow and movement blur is not a problem, use the normal lighting method, and when the inspection speed is fast and movement blur is a problem, use a strobe light synchronized with the positioning signal to illuminate the moving image. All you have to do is freeze it and read the pattern.

FIG. 6 shows the configuration of the sensor 8 for determining the area distribution of the inspection pattern. The sensor includes a plurality of light-receiving zones 7 arranged concentrically around the origin.

In this sensor, for example, as shown in FIG. 7, photoelectric conversion elements such as solar cells are arranged concentrically as light-receiving zones 7 with isolation zones 9 provided for the purpose of separating each light-receiving zone 7. A photoelectric sensor that does this can be considered. Alternatively, it is also possible to use an industrial television camera or a two-dimensional solid-state image sensor as the sensor. When using an industrial television camera or two-dimensional solid-state image sensor, calculate the value of √ ² + ² corresponding to the distance from the origin at address (i, j) of the inspection pattern as shown in Figure 8. By doing this (i, j)
It is determined which preset light-receiving zone 7 the address belongs to, and the density information of each light-receiving zone 7 is totaled, or in the case of a binary image, the number of light and dark information of each light-receiving zone is accumulated and the area is calculated. You can find the distribution. This arithmetic processing can be performed by a computer.

By using such a sensor 8, the area distribution of the inspection pattern as shown in FIG. 9 can be determined. Here, the number and width of the light-receiving bands 7 of the sensor 8 are set to values that can adequately represent the pattern to be inspected.

In the present invention, an example of identifying an inspection pattern or inspecting a defect by detecting the area distribution using the photoelectric sensor shown in FIG. 7 will be described below. FIG. 10 shows a block diagram of signal processing. It is assumed that the object to be inspected 6 is set in the inspection zone 5 of the inspection station 3 in FIG. 4 or 5. The light from the lamp 14 is irradiated onto the object 6 to be inspected and reflected there, and the inspection pattern 10 is imaged through the lens 12 on the photoelectric sensor 8 having light receiving zones 7 arranged concentrically. The electrical output corresponding to the intensity of the incident light converted by the photoelectric conversion element of the light-receiving zone 7 is amplified by the preamplifier 15 corresponding to each light-receiving zone 7 . The output of this preamplifier 15 represents the area distribution of this inspection pattern 10.

Next, using the signal processing device 16, an inspection pattern is identified or a defect inspection is performed from the obtained area distribution pattern. The signal processing device may be constructed using a microcomputer 19 as shown in FIG. 11, for example. Preamplifier 1
5 is converted into a digital quantity by an analog-to-digital converter 18 via a multiplexer 17 and stored in the memory of a microcomputer 19. Here, the signal a is a positioning signal indicating that the inspection pattern 10 is set in the inspection zone, and the signal b is a signal from a photoelectric switch or the like that detects the presence or absence of the object 6 to be inspected in the inspection zone. signal a
When signal b is turned on and it is notified that positioning has been completed, and when it is confirmed that the object 6 to be inspected is present with the signal b turned on at this time, the area distribution data is read. Next, using the read data, the degree of consistency between the area distribution of the inspection pattern and the area distribution of the previously registered reference pattern is determined. For example, simple similarity S is often used as a quantity representing this degree of matching. The microcomputer 1 calculates this S.
9. Here, this simple similarity S
An example in which inspection pattern identification or defect inspection is performed by determining . As shown in FIG. 12, the simple similarity S is calculated from the element P(i) of the area distribution of the test pattern and the element Q(i) of the area distribution of the reference pattern. It can be found using the formula. Here, i indicates the number of each light receiving zone. As mentioned earlier, since the area distribution is a one-dimensional pattern, calculation of S is also much simpler than in a two-dimensional pattern.

S is 1 if P(i) and Q(i) are exactly the same pattern, 0 if there is no correlation at all, and a value between 1 and 0 if the similarity is in between. . In other words, the higher the degree of similarity, the closer the value is to 1. Therefore, by using this value of S, inspection pattern identification or defect inspection can be performed. Here, identification means identifying which of the reference patterns the test pattern is by determining which of the reference patterns registered in advance the test pattern shows the highest degree of similarity. In addition, for defect inspection, there is only one type of pattern flowing through the production line, and the same type of pattern is registered as a reference pattern, and the simple similarity is calculated, for example, 0.8≦S≦1.0 → Good product 0≦S<0.8 →This can be done by determining that the product is defective.

In the above explanation, when calculating the simple similarity S, it is not necessarily necessary to select the numbers (i) of all the light-receiving zones 7. Accuracy can be improved by finding out in advance the number (i) of the light-receiving zone that best represents the characteristics of the pattern, and calculating S using the area distribution corresponding to the output of this light-receiving zone. It is also effective to add the following method to further increase reliability. In other words, the pattern of the area distribution is considered to be more important than its absolute value. For example, the absolute value of the area distribution changes when the light source intensity fluctuates, but the pattern does not change. Therefore, as shown in Figure 13, in order to standardize the area distribution, the ratio fi/〓 _i fi of the elements of the area distribution is calculated, and if this is newly adopted as the area distribution, an inspection system that is resistant to changes in the environment can be created. It can be done.

This invention defines an area distribution, which is an attribute value that does not change with respect to the rotation of a figure, and uses this area distribution as an identification parameter to identify patterns or defect inspection of an object to be inspected that has a degree of freedom in the rotational direction. Easy to implement. In addition, since information is compressed by converting a two-dimensional pattern into a one-dimensional pattern called area distribution, the subsequent identification process can be done with a simple process. very effective.

[Brief explanation of the drawing]

Fig. 1 is a flowchart showing the state of the object to be inspected, Fig. 2 is a pattern diagram to explain the area distribution, Fig. 3 is a graph showing the relationship between the radius and brightness state in Fig. 2, Fig. 4 and Figure 5 is a configuration diagram of the inspection station, Figures 6 and 7 are configuration diagrams of the sensor, Figure 8 is a diagram explaining the method for determining area distribution, and Figure 9 shows the relationship between radius and area distribution. 10 and 11 are block diagrams of signal processing, FIG. 12 is a graph explaining simple similarity, and FIG. 13 is a graph for normalizing area distribution. DESCRIPTION OF SYMBOLS 1... Manufacturing equipment, 2... Conveyor, 3... Inspection station, 4... Setting section, 5... Inspection zone, 6... Inspection target object, 7... Light receiving zone, 8...
Analog sensor, 9... Isolation zone, 10... Inspection pattern, 12... Lens, 1
4...Lamp, 15...Preamplifier, 16...Signal processing device.

Claims (1)

[Claims] 1 The pattern marked on the object to be inspected is divided into a plurality of concentric bands, and the distribution of the values is determined by detecting the value corresponding to the optical quantity in each concentric band of the pattern, while the reference The pattern is divided into a plurality of concentric bands, and the value distribution is determined by detecting the value corresponding to the optical quantity in each concentric band of the reference pattern, and the distribution of the values is determined by dividing the pattern into a plurality of concentric bands. A pattern in which the distribution of optical quantities in the concentric bands of the pattern on the object to be inspected is compared with the distribution of optical quantities in the concentric bands of the reference pattern, with the centers of the pattern and the reference pattern being aligned. Inspection method.