JPH10123066A - Apparatus and method for detection of singularity point - Google Patents

Abstract

PROBLEM TO BE SOLVED: To make it possible to judge kind of defect by detecting a singularity point existing in an object to be inspected with high accuracy by an image processing operation. SOLUTION: A take-up roll 2 winds up a sheet 1 continuously. An illumination lamp 3 is used in such a way that the sheet 1 is irradiated with light, and its reflected light is received by a CCD camera 4. A noise is removed from input image information, a maximum value and a minimum value in every range (a small range) which is composed of five longitudinal pixels and five transverse pickles are detected, their difference is calculated so as to be compared with a threshold value, and a bit map is created inside a storage circuit. On the basis of the bit map or the like, data is analyzed, and the kind, the degree and the like of a singularity point are detected automatically.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an apparatus and a method for detecting a singular point existing on an object to be inspected with high accuracy by image processing and determining the type of the defect. The present invention relates to a singularity detecting device and method capable of detecting whether the change of the singularity is slow or rapid.

2. Description of the Related Art Conventionally, CCD (charge coupled device)
Many automatic visual inspection techniques using image processing using a camera are known. Most of them detect a peculiar part by differentiating the image data, or detect a peculiar part by comparing the illuminance of the image data with a certain threshold value.

[0002]

However, these automatic appearance inspection techniques may have the following disadvantages. 1. In the case of detecting a singular part by differential processing, a change in the difference between adjacent parts is calculated, so that only a part where the illuminance changes abruptly is detected as abnormal, and any gradual change cannot be detected. For example, sheet 1 as shown in FIG.
It is assumed that a distinctive unique portion 10 and an unclear distinctive portion 11 exist on the surface of. Here, the illuminance is represented by 256 gradations. FIG. 26 is a graph showing illuminance data between A1 and A2 (part of the distinctive unique portion 10) in FIG. FIG. 27 is a graph showing the illuminance data between B1 and B2 (a part of the unclear portion 11) in FIG. Comparing the distinct singular part 10 with the unclear singular part 11, the conventional method (differential method between two pixels) detects the singular part 10 because the differential value is large (that is, the change rate of the illuminance is large). Detection is possible,
There is a possibility that the detection of the unclear singular part 11 becomes impossible because the differential value is small.

[0003] 2. When detecting a singular part by differential processing, only a change in a certain linear direction (one of vertical, horizontal, and oblique) is detected, and no change is detected in any direction (for example, as a surface). 3. When detecting an unusual part of the illuminance of the image data at a certain threshold value, the nature of the background of the inspection surface varies (up and down variations in the overall illuminance due to changes in illumination, etc., or gradual changes in unevenness). In some cases) the inspection is difficult. 4. There are few methods for automatically classifying the type of a unique part such as a detected defect based on its shape. SUMMARY OF THE INVENTION An object of the present invention is to solve the above-mentioned drawbacks of the prior art, and to extract a unique part such as a defect of an object to be inspected with high sensitivity and reproducibility. A new singularity detecting apparatus and method for classifying even the types of defects.

[0004]

In order to achieve the above object, according to the first aspect of the present invention, there is provided a sensor for converting predetermined information of an object to be inspected into image information, and a sensor for converting the predetermined information into image information. Storage means for storing the image information obtained by dividing the image information into a plurality of small areas each having a predetermined number of pixels in each of the vertical and horizontal directions and being shifted by one pixel in the vertical and / or horizontal direction. The maximum / minimum value detecting means for detecting the maximum value and the minimum value of the image information in each small area by sequentially comparing the size of the image information for each pixel for each pixel, and the maximum / minimum value detecting means Error calculating means for calculating an error between the maximum value and the minimum value; comparing the error calculated by the error calculating means with a predetermined threshold value; Judge that a singular point exists in the test object It was constructed with a constant section.

Here, the sensors are, for example, optical sensors, temperature sensors, infrared sensors, X-ray sensors, ultrasonic sensors, ultraviolet sensors, nuclear magnetic resonance (MRI) sensors, lasers, and the like. All sensors that can convert various types of information into image information. Also,
A method of setting a plurality of small regions (regions composed of one or more pixels in the vertical direction and one or more pixels in the horizontal direction), finding a maximum value and a minimum value among them, and calculating an error thereof. Therefore, instead of calculating the change in displacement at minute intervals on a straight line only in the horizontal, vertical, or oblique direction as in the conventional differential method, the change in displacement in a plane called a small area is calculated. The degree of freedom and redundancy can be given to the calculation method of the error for the calculation. For this reason, it is possible to easily detect even a singular point having a gradual displacement which could not be detected conventionally. In addition, by changing the size of the small area in accordance with the type and size of the singular point to be detected, the degree of the singular point, the color, and the like, it is possible to cope with various uses and fields that are difficult to obtain by the conventional differential processing.

The error calculated by the error calculating means for each small area is compared with a predetermined threshold value. Can be determined to exist. Here, it is also possible to determine only whether or not a singular point exists, for example, by comparing the error between the maximum value and the minimum value with a threshold value and detecting the singular point. The difference between the maximum value and the minimum value for each small area is compared with a predetermined threshold value, and a specific numerical value or symbol (for example, 1) is set to a value smaller than the threshold value. In the case of, another specific numerical value or symbol (for example, 0)
It is also possible to create a bitmap allocated to pixels at a predetermined position (for example, the center of the small area) of each of the small areas. In this case, the position, size, shape, and the like of the singular point can be determined by visually observing the created bitmap. It is also possible to create a bitmap in which the threshold value has a plurality of levels and a predetermined numerical value (for example, a numerical value such as 1, 2, 3,..., For each level) is assigned to each level. In such a case, the state of the singular point, such as a contour line, can be determined including its degree.

[0007] The object to be inspected is usually confusing with a singular point, and patterns and marks (signs), various stains and distortions, unevenness in production, differences in gloss, differences in gloss, In many cases, there are variations in illumination from illumination, fixed variations among elements of light receiving pixels of a camera, and the like (hereinafter, referred to as fixed noise). In order to remove this fixed noise, as in the invention according to claim 3, the storage means includes a first storage circuit for storing at least image information of an inspection master having no singular point converted by the sensor in advance. A second storage circuit for storing the image information of the inspection object converted by the sensor again, and a corresponding one of the pixels of the first storage circuit based on the image information of each pixel of the second storage circuit. Fixed noise removing means for storing the remainder after subtracting image information as image information may be provided between the storage means and the maximum / minimum value detecting means. In this case, first, image information of the inspection master having no singular point is fetched from the sensor and stored in the first storage circuit. After that, the image information of the inspection object is similarly captured from the sensor and stored in the second storage circuit. Then, the image information of each pixel of the corresponding first storage circuit is subtracted from the image information of each pixel of the second storage circuit, and the rest is stored as image information. As a result, it is possible to detect only the singular points existing in the test object in a form in which the information is removed as fixed noise without being influenced by various information originally possessed by the inspection master and inherent variations existing in the sensor. . The type of fixed noise is limited to some extent when detecting singular points of the test object in-line as described above, but when detecting singular points off-line, various patterns, shapes, etc. are used as fixed noise. Can be considered.

Although the elimination of the fixed noise has a considerable effect on the detection of a singular point, it is possible to eliminate minute noise for each light-receiving pixel data generated randomly in each pixel. As described in the fourth aspect, the image information stored in the storage means or the image information processed by the fixed noise elimination means is a small area having a predetermined number of pixels in each of the vertical and horizontal directions, one pixel at a time in the vertical direction and / or It is divided into a plurality of small areas shifted in the horizontal direction, the sum of image information of each pixel is obtained for each of the small areas, divided by the number of pixels constituting the small area, and an integer part is predetermined in the small area. Random noise elimination means assigned to pixels at predetermined positions may be provided between the storage means or the fixed noise elimination means and the maximum / minimum value detection means. First, the average value of the image information in the small area is obtained by calculating the sum of the image information of each pixel in each small area and dividing the sum by the number of pixels constituting the small area. By averaging and rounding down decimal places, random noise that occurs suddenly in pixel units can be removed. The data after the removal is allocated to a pixel at a predetermined position (for example, the center of the small area) in the small area. By repeating such operations, random noise of the entire image can be removed.

In each of the above-described processes, first, after image information is acquired by the sensor, fixed noise removing means, random noise removing means, maximum / minimum value detecting means,
It is also possible to perform operations of the error calculating unit, the determining unit, and the bitmap creating unit, and to continue the operations in series, such that the image information is obtained again by the sensor after the completion of the arithmetic processing. However, as in the invention according to claim 5, the storage means is divided into a plurality of storage circuits, and the image information converted by the sensor is stored in one storage circuit in parallel with another storage circuit. One of which performs the respective arithmetic processing of the maximum / minimum value detecting means, the error calculating means, the determining means, the bitmap creating means, and the random noise removing means, and performs the arithmetic processing and the storing processing in the storage circuit. After at least one of the processes is completed, the one storage circuit calculates each of the maximum and minimum value detection means, the error calculation means, the determination means, the bitmap creation means, and the random noise removal means for the stored image information. A plurality of storage circuits are sequentially stored, such as performing processing and performing processing for storing image information converted by the sensor in one of the other storage circuits. It is also possible to simultaneously process the arithmetic processing and the alternate storage while changing the role arithmetic processing and the a. As described above, by simultaneously performing the acquisition of the image information by the sensor and the respective arithmetic processings, it is possible to easily cope with continuous operations such as in-line. Also,
The processing speed can also be increased. Further, depending on the speed of the arithmetic processing, the amount of processing, and the like, the capacity of the storage circuit can be limited and the storage circuit can be divided into a plurality of storage circuits.

As described above, various sensors can be applied to the sensor of the singularity detecting device. As described in the sixth aspect of the present invention, the sensor is a device that reflects or transmits reflected light from illumination to the object to be inspected. It is also possible to comprise at least one or more optical cameras that convert predetermined information of light into image information. When the inspected object is an opaque or translucent object, a singular point on the outer surface of the inspected object can be detected by reflected light from the illumination, and when the inspected object is a transparent or translucent object , It is possible to detect a singular point by transmitted light from illumination. Further, in the case of a translucent body, transmitted light can be used, and in the case of a transparent body, reflected light can be used. The optical cameras may be arranged continuously in a line or may be arranged vertically in several stages.

Further, when the specular reflection light from the illumination is received by the optical camera with respect to the inspection object, it is possible to obtain a strong irradiation by shortening the distance between the illumination and the inspection object. However, as in the invention according to claim 7, the sensor receives specularly reflected light from illumination of the inspection object, and the distance between the illumination and the inspection object is a distance between the sensor and the inspection object. And more than 300 mm apart. In this regard, by setting such a distance, it can be expected that the light from the illumination will be parallel light to some extent, and the detection of the singular point will be more accurate than when the distance between the illumination and the test object is narrowed and strong irradiation is obtained. Well possible. The reason for reducing the distance between the illumination and the test object is that it is expected that diffuse reflection will increase in exchange for obtaining strong irradiation.

In addition, even when reflected light or transmitted light from the illumination of the object to be inspected is directly received, the singular point can be sufficiently detected. A grid having a plurality of slits may be provided between the test objects. The width of the slit can be changed according to the size of the singular point to be detected. Due to the presence of the slit, the received image information itself has alternating bright and dark streaks.However, if there is a singular point in the test object, parallel rays are irregularly reflected at that point, and The irregularly reflected light due to the singular point clearly emerges brightly. On the contrary, the singular point may appear dark on the streak of the bright part. In such a case, the singular point has a significantly different illuminance than the surrounding illuminance, so that the detection accuracy of the singular point is improved. The bright and dark streaks can be removed in advance by the above-described fixed noise removing means. The shape of the slit is free, such as parallel or mesh.

The singular point can be sufficiently detected even if the threshold value of the determination means or the bitmap creation means is a constant value. The threshold value of the bitmap creating means is at least one row in a predetermined direction according to at least one of aberration of the lens itself built in the optical camera and inherent variation of each lens when a plurality of optical cameras are used. Can be set in advance so as to have a different threshold value for each pixel or a different threshold value for a plurality of pixels. Since the lens usually has a curved surface, its central portion tends to be bright and becomes darker toward its peripheral portion. Further, when a plurality of lenses are used, there are manufacturing variations and the like. For this reason, the lens-specific correction is applied to the threshold value, and a different threshold value is set for each pixel or a stepwise different threshold value is set for each portion including a plurality of pixels. The predetermined direction is not limited to the width direction of the object to be inspected, but includes a free direction, and also includes a case where a plurality of optical cameras are vertically arranged.

The type of the defect at the singular point can be determined by visually observing the bitmap created in claim 2, but the bitmap created by the bitmap creating means as in the invention according to claim 10 is provided. A totaling means for sequentially totaling one number for each row or each column in at least one direction of the vertical direction and the horizontal direction, and a projection distribution obtained by graphing the totaling result of the totaling means with the total number as a height direction. Creating means, at least the width of the unique portion in each of the vertical and horizontal directions based on the projection distribution creating means, the ratio of the width in the vertical direction to the width in the horizontal direction, the area of the unique portion graphed, the width in the vertical direction or The ratio divided by the width in the horizontal direction, the number of irregularities in the height direction of the singular part graphed, the image information of each pixel after each arithmetic processing by the fixed noise elimination means or the random noise elimination means At least one of the items including the determination as to whether the large value belongs to a plurality of predetermined setting ranges is compared with a reference value predetermined for each type of defect, and the degree of coincidence is determined. The defect type judging means for judging the defect having the largest defect number or the defect having the highest degree of coincidence after performing the weighting process for each item as the defect type may be further provided. In the bitmap, a pattern of a unique portion is drawn with a specific numerical value or symbol (for example, 0 and 1 (the numerical value is given as an example and is not limited to this)). Using a computer or the like, it is automatically determined to what kind of character this pattern belongs. For this reason, the features of the unique portion are aggregated into several pieces of information such as the width and the area of the height and width, and this data is compared with a reference value determined in advance for each type of the defect, so that the type of the defect and the position of the defect, The degree and the like can be automatically determined instantaneously. The defect with the highest degree of coincidence may be determined as the defect type without assigning any weight to each item, but the weighting process is performed for each item and the defect with the highest degree of coincidence in the overall result is determined. May be determined as the defect type.

In this way, it is possible to determine the type of defect, and it is possible to regard all defects as defects irrespective of the size of the unique portion. When the defect type is determined by the defect type determination unit, the width of the unique portion in the vertical and horizontal directions based on the projection distribution generation unit is compared with a pass / fail determination width determined in advance for each defect type, and the vertical and horizontal directions are determined. It is also possible to further comprise a defect judging means for judging a defect when at least one of them is larger than the pass / fail judgment width. With this, it is possible to judge whether the defect is acceptable or not depending on the size of the defect.

[0016] The category is different, and predetermined information having reflected light or transmitted light from the illumination of the object to be inspected is converted into image information by an optical camera, and the image information is converted. Is stored in a storage means in a bitmap form for each pixel, and the image information is divided into a plurality of small areas each having a predetermined number of pixels in the vertical and horizontal directions and shifted one pixel at a time in the vertical and / or horizontal direction. Then, the maximum value and the minimum value of the image information in each small region are detected by successively comparing the size of the image information for each pixel in each of the small regions, and an error between the maximum value and the minimum value is detected. After the calculation, the calculated error is compared with a predetermined threshold value, and when the threshold value is exceeded in at least one or more small areas, it can be determined that a singular point exists in the test object.

[0017]

Embodiments of the present invention will be described below with reference to the drawings. In FIG. 1 showing the first embodiment of the present invention, a plain sheet 1 is obtained by coating a surface of a white paper as a raw material. Take-up roll 2
Are designed to continuously wind the sheet 1. Lighting 3
Is for irradiating the sheet 1 with light, and the reflected light is received by the CCD camera 4. The CCD camera 4 corresponds to an optical camera.

Next, the operation will be described. Among various inspections required for quality control in a manufacturing process by converting or the like, appearance inspection is one of important processes. According to the present invention, in order to automate the appearance inspection, detection of a unique portion such as a defect after input by the CCD camera 4 is performed by image processing. As shown in FIG. 1, before winding on the winding roll 2, defects on the surface of the sheet 1 (for example, scratches, dirt, foreign matter,
The process of visually inspecting for unevenness of the coating, etc.) by the reflected light is automated by the present invention. In particular, the present invention is suitable for performing continuous inspection in-line.

The operation process will be described below. 1: Process 1-1 (Process for capturing image data from CCD camera 4) A 600 mm wide CCD line camera having 4096 horizontal light receiving pixels was used to inspect a portion having a width of 600 mm. The inspection sheet 1 was set at a speed of 0.1 m / sec, and the shutter speed of the CCD camera 4 was set at 1.5 msec (hereinafter referred to as ms). Illumination 3, which is the light source that illuminates the surface to be inspected
Uses a high-frequency fluorescent lamp with a width of 700 mm. The illumination 3 is 500 mm away from the sheet 1 to be inspected, and the CCD camera 4 is 2 m away from the sheet 1 to be inspected.
The camera was installed so that the reflected light of the illumination on the surface to be inspected was specularly reflected (the incident angle and the reflection angle were equal) to the camera. 4096 horizontal
The data of the light receiving pixels of the CCD camera 4 in one vertical line is CC
Storage circuit 1 which is output from D camera 4 every 1.5 ms
(The storage circuit is divided into halves and referred to as a storage circuit 1 and a storage circuit 2). 4096 horizontal memory circuits, 2 vertical memory circuits
When the number of columns becomes 56, the memory for storing the light receiving pixel data from the CCD camera 4 is switched to the storage circuit 2. Thereafter, as described above, the light receiving pixel data from the camera is alternately switched and stored in the storage circuits 1 and 2 so that one half of the storage circuit can store the data from the CCD camera 4.
While the data is being stored, half of the other storage circuits perform a process of detecting a unique portion such as a defect. By this method, CCD
Data from the camera 4 can be continuously taken into the storage circuit, while detection of a unique portion such as a defect can be continued. FIG. 2 shows image data in the vicinity of the unclear peculiar part 11 and FIG. 8 shows image data in the vicinity of the peculiar peculiar part 10.

2: Processing 1-2 (Fixed Noise Elimination Processing) In storing data of the light receiving pixels from the CCD camera 4 in a storage circuit (including the storage circuits 1 and 2 and corresponding to storage means). Since there is a fixed variation (hereinafter referred to as a fixed noise) between the elements of the light receiving pixels of the CCD camera 4, in order to correct this, a state (inspection master) in which there is no peculiar part such as a defect on the inspection surface is set in advance. 1 horizontal line (40 horizontal)
(96 pieces, 1 piece) data is captured by the CCD camera 4, and this is stored and stored in another storage circuit 3 of 4096 pieces of 1 piece, and the CCD camera 4 during the actual inspection.
The difference between the data (4096 horizontal, one vertical) output for each line output from the memory circuit 3 is used. Such processing corresponds to a fixed noise removing unit.
FIG. 3 shows image data in the vicinity of the unclear peculiar part 11 and FIG. 9 shows image data in the vicinity of the peculiar peculiar part 10 in a part of the storage circuit after the fixed noise removal processing.

3: Processing 1-3 (random noise removal processing) Next, when detecting a unique portion such as a defect, the light receiving sensitivity and the element sensitivity of each pixel of 4096 pixels horizontally and 256 columns vertically stored in the storage circuit are detected. In order to eliminate minute irregularities (hereinafter referred to as random noise) for each light receiving pixel data of the CCD camera 4 which are randomly generated due to variations or the like, a matrix of, for example, 3 pixels vertically and 3 pixels horizontally is set. The random noise is removed by replacing the numerical value of the central pixel in the matrix with the simple average value of all nine pixels while shifting one pixel at a time in the vertical and horizontal directions. Here, the decimal part of the average value was rounded down. However, replacement is not limited to the center pixel. Such processing corresponds to a random noise removing unit. FIG. 4 shows image data in the vicinity of the unclear peculiar part 11 and FIG. 10 shows image data in the vicinity of the peculiar peculiar part 10 in a part of the storage circuit after the random noise removal processing. In this processing, a dropout of two pixel lines occurs in the vertical direction and the horizontal direction at the end of the arithmetic processing portion. In the present system, in order to inspect the entire surface of the inspected portion, pixels to be taken into the memory are processed by four lines larger than the pixels requiring analysis, thereby preventing omission.

4: Processing 1-4 (processing of replacing the difference between the maximum value and the minimum value for every 5 * 5 pixels) In the storage circuit of FIG. 4 or FIG. 10, for example, a range of 5 pixels vertically and 5 pixels horizontally ( The maximum value and the minimum value are detected (corresponding to the maximum / minimum value detecting means) and the difference between them is calculated (corresponding to the error calculating means) for each of the sub-areas (corresponding to the small area). Write the data to another relevant storage circuit.
This processing is performed while shifting all pixels in the storage circuit vertically and horizontally for each pixel. FIG. 5 shows image data in the vicinity of the unclear singular part 11 in a part of the storage circuit after this processing.
FIG. 11 shows image data in the vicinity of the distinctive unique portion 10.

5: Processing 1-5 (Binarization Processing) For the storage circuit (FIGS. 5 and 11), for example, 1 is converted into 5 when the number is 5 or more, and 0 is converted when the number is less than 5. (Hereinafter, the value of 5 is referred to as a threshold value, and this process is hereinafter referred to as a binarization process.) FIG. 6 shows a singular portion 11 in which the result of performing the binarization process on all the pixels of the storage circuit is unclear. FIG. 12 shows a clear unique part 11. FIGS. 6 and 12 correspond to a bitmap, and the processing of the processing 1-5 corresponds to a bitmap creating unit. However, when two or more CCD cameras 4 are used, variations in detection of unique parts (due to variations in focus setting, etc.) occur in each CCD camera 4 and when the field of view of the CCD cameras 4 is widely used. Means that the CCD camera 4
The illuminance data captured at the center of the CCD camera 4 has a problem that the central portion is bright and the outer portion is dark. In this embodiment, since two CCD cameras 4 are used, the threshold value for performing the binarization process needs to be different for each of the two CCD cameras 4. Further, even in the same CCD camera 4, the inspection processing unit is divided into five parts in the arrangement direction of the CCD line camera from the outer part which is one end of the lens to the center part of the lens, and further to the outer part which is the other end of the lens. Thus, the above-described problem is solved by setting a threshold value for each processing unit. The threshold of the present embodiment is the first CC
In the D camera 4, 3, 4, 5, 4, 3 (the center of the field of view of the CCD camera 4 is 5).
3, 4, 3, and 3 (the center of the field of view of the CCD camera 4 is 4), but the case of the threshold value 5 will be described below by taking the center of the first CCD camera 4 as an example for simplification.

6: Processing 1-6 (Projection Distribution Processing) The number of data 1 is totaled for each column and each row in the vertical and horizontal directions of the binarized storage circuit (corresponding to totaling means). . The results of the aggregation are displayed in a graph as shown in FIG. 7 for the unclear unique portion 11 and FIG. 13 for the clear unique portion 11. Such graphing corresponds to a projection distribution creating unit.

7: Process 1-7 (data of a unique part such as a defect)
Data analysis processing) From the projection distribution data, the width, vertical / horizontal width, area, area / horizontal width, and the number of irregularities of the unique portion are calculated, and the data of the storage circuit (FIGS. 3 and 9) are further calculated. Calculate the maximum image value and the intermediate image value (described later) of the unique portion from the data. The characteristic of the unique part is analyzed based on the calculation result. In this analysis data, the method of calculating the maximum image value and the intermediate image value will be described. First, the threshold value of the maximum image value is set to 20 and the threshold value of the intermediate image value is set to a range of 10 to 19. Set in advance. When the processing in FIG. 3 is performed, the numerical values are compared line by line, and since there is no line having the maximum value of 20 or more, the value of the maximum image is set to 0. The value is 1. Similarly, in FIG. 9, when the maximum value is viewed for each line while shifting the horizontal line of the peculiar portion one line at a time in the vertical direction, the maximum value is 20 or more, and the maximum value is in the range of 10 to 19. It can be seen that there is also a line. Therefore, the maximum image value is 1 and the intermediate image value is also 1.

8: Processing 1-8 (Defect Type Judgment Processing for Unique Part) A classification data table (Table 1) in which characteristics of each defect type are analyzed in advance is prepared. The defect having the highest degree of matching of the feature is determined as the defect type by matching the feature with the unique portion.

[0027]

[Table 1] In the example of the unique portion in FIG. 2, the horizontal dimension is 3.6 mm (24 pixels * 0.15 mm / pixel), and the vertical dimension is 3.15 mm (2 pixels).
1 pixel * 0.15 mm / pixel), number of irregularities = 2, vertical dimension / horizontal dimension = 0.87, area / horizontal width = 15.4, maximum value of image excluding fixed error = 15. Compare with Table 1

[0028]

[Table 2] Therefore, it is determined that the defect has the largest total score. FIG.
In the example of the peculiar part of the above, since the image data is 0.15 mm / pixel, the horizontal dimension is 2.85 mm (19 pixels * 0.15 m
m / pixel), vertical dimension 3.75 mm (25 pixels * 0.15)
mm / pixel), number of irregularities = 1, aspect ratio = 1.32,
In the area / width ratio = 12.6, the maximum value = 22, and the maximum value for each line, there are 18 and 15 intermediate values. Looking at Table 1,

[0029]

[Table 3] Therefore, it is determined that the crack has the largest number of total points. In this embodiment, each item of the unique portion is evaluated equally, but each item may be weighted in order to make the characteristic of the unique portion to be detected remarkable. The above processing corresponds to a defect type determination unit.

9: Processing 1-9 (Defect Pass / Fail Judgment Processing) The pass / fail judgment dimensions of the vertical width and the horizontal width for each defect type (automatically convert defective pixel dimensions to mm dimensions) Table 4 are prepared in advance.

[0031]

[Table 4] 7: If one or both of the vertical dimension and the horizontal dimension calculated in the processing 1-7 are larger than the pass / fail dimensions in Table 4, it is determined that the defect is not a defect. Such processing corresponds to a defect determination unit. In the example of FIG. 2, the minimum defect size of the missing defect in Table 4 is 3 mm in the vertical dimension and 3 mm in the horizontal dimension.
mm, vertical dimension = 3.15 mm, horizontal dimension = 3.6
mm, it is determined to be a defect. In the example of FIG. 8, the minimum defect size of the crack defect shown in Table 4 is 5 mm in vertical dimension and 1 mm in horizontal dimension.
Is 3.75 mm in length and 2.85 mm in width, so it is determined to be a defect. Experiments have shown that the distance between the illumination 3 and the sheet 1 is at least 1/5 of the distance between the CCD camera 4 and the sheet 1 and that a singular point can be detected with high accuracy if the distance is at least 300 mm. Has been confirmed. It is presumed that narrowing the distance between the illumination 3 and the sheet 1 increases the irregular reflection on the surface of the sheet 1 at the cost of obtaining a strong irradiation, and conversely lowers the detection accuracy of the singular point.

Next, a second embodiment of the present invention will be described. The description of the same elements as those in FIG. 1 will be omitted.
In FIG. 14 showing the second embodiment of the present invention, an illumination 3 is provided with a CCD camera 4 with a transparent or translucent sheet 1 interposed therebetween.
And is facing. A grid 5 is provided between the lighting 3 and the sheet 1.
Is placed. The grid 5 is provided with a plurality of slits 6 in parallel as shown in FIG.

Next, the operation will be described. In the first embodiment,
Although the reflected light for the sheet 1 is input to the CCD camera 4, the transmitted light for the sheet 1 is
Input to the camera 4. The processing after the input of the image information is the same as in the first embodiment. However, because of the transmitted light, it is possible to detect not only information from the outer surface of the test object but also singular points including defects and the like existing inside. Further, by placing a grid 5 between the illumination 3 and the sheet 1, an image having light and dark in a grid shape can be obtained by C
It will be imported to the CD camera 4. However, such bright and dark images can be removed as fixed noise as described above. In this state, if there is any defect or the like in the sheet 1, the defect can be clearly highlighted in a bright and dark image due to irregular reflection from the unique portion. The interval between the slits 6 is changed in accordance with the size of a singular point to be detected. In this embodiment, a slit interval of about 1 mm is applied to detect a defect of about 1 mm. The height of the slit is 20
Mm.

In such a case, the singular point has a significantly different illuminance than the surrounding illuminance, so that the detection accuracy of the singular point is improved. That is, even if a unique portion such as a defect is a unique portion of a plain transparent or opaque coated surface that is indistinct and difficult to detect under conventional optical conditions, the change in illuminance is brought out as a clear change in illuminance. Can be done. Further, the present invention is not limited to the lattice 5 having a plurality of slits 6 in parallel,
Considering that the bright and dark images can be removed as fixed noise, a grid having a mesh-like pattern may be used. In addition, in order to confirm the usefulness of the embodiment of the present invention,
Similar processing was performed on the unique portions 10 and 11 of the same sheet 1 by the conventional method (differential method between two horizontal pixels), and the results were compared with the results of the present invention.

1: Process 2-1 (Process for Importing Image Data from CCD Camera 4) The process is the same as process 1-1. A part of the data after the fetch is the same value as in FIGS. 2 and 8. 2: Process 2-2 (Fixed Noise Removal Process) The fixed noise removal process of the storage circuit is the same as the process 1-2. Some of the data of the storage circuit after the processing are shown in FIGS.
And the values are as shown in FIGS. 16 and 21. 3: Processing 2-3 (random noise elimination processing) In all the pixels in FIGS. 16 and 21 of the storage circuit, the average value among the three pixels in the horizontal direction is substituted for the central pixel of each of the three pixels, and the processing is performed in all directions. To process. FIG. 17 shows image data in the vicinity of the unclear peculiar part 11 and FIG. 22 shows image data in the vicinity of the peculiar peculiar part 10 in a part of the storage circuit after the processing. In addition, in the case of this processing, since two lines are missing in the horizontal direction, it is necessary to increase the processing range accordingly. No vertical dropout occurs.

4: Processing 2-4 (differential processing) For all the pixels shown in FIGS. 17 and 22 of the storage circuit, the difference between the two horizontal pixels is calculated while shifting one pixel at a time vertically and horizontally, and the difference is substituted for the left side of the two pixels. The data is stored at the corresponding position in the storage circuit.
FIGS. 18 and 23 show a part of the storage circuit after the processing.

5: Processing 2-5 (Binarization Processing) For all the pixels in FIGS. 18 and 23, a part of the storage circuit after the binarization processing with the threshold value set to 5 is shown in FIGS. 24. Referring to FIG. 24, it can be seen that a peculiar portion is detected, but that of FIG. 19 is not detected at all. FIG.
In FIG. 19, even if the threshold value is changed to 5, 4, 3, or 2 for binarization, the result is the same as in FIG.
In FIG. 18, FIG. 20 shows the result of performing the binarization processing with the threshold value set to 1. Referring to FIG. 20, it can be seen that, in addition to the singular part, the random noise part is also binarized to become a singular part. From the above comparison results, it is understood that it is difficult to detect a singular part that changes slowly with the conventional differentiation method.

Next, the following was performed to confirm the effects of the optical system of the present invention. A. In the optical system of the above embodiment, a unique portion of the coated surface on the opaque sheet 1 was inspected using reflected light. B. In the optical system of the above embodiment, a unique portion of the coating surface on the transparent sheet 1 was inspected using transmitted light. C. In A, the illumination distance from the surface to be inspected is 500 mm
To 200 mm. D. In B, no grating 5 was used. E. FIG. In A, the threshold values of the inspection processing units of the first CCD camera 4 are all 5, and the threshold values of the inspection processing units of the second CCD camera 4 are all 4. F. In A, the threshold values of the inspection processing units in the two CCD cameras 4 were all set to 5. At the time of inspections A to F, unique parts such as defects were visually observed in parallel to judge pass / fail. Here, when the inspection device determined that a peculiar portion that was not a defect was a defect, it was determined to be a phantom, and when a defect was detected as a defect, the occurrence frequency was calculated as detection. The detection rate is expressed by a number obtained by dividing the number of defects detected by the inspection machine that are equivalent to defects in the visual inspection by the number of visual defects and multiplying by 100, and the illusion rate is visually detected by the inspection machine. In the classification, the number of those corresponding to the pass was divided by the number of visual defects to obtain 10
It was represented by a number multiplied by 0. Table 5 shows the results.

[0039]

[Table 5] It is clear from the table that the illusions of A and B are low and the detection rate is high.

[0040]

As described above, the present invention (claim 1)
According to the configuration described above, the maximum value and the minimum value are detected for each small area in the image information converted by the sensor, and the error between them is compared with a threshold value. In this way, it is possible to easily detect not only a portion where the displacement changes abruptly but also a singular point where the displacement which cannot be detected by the conventional differential processing is arbitrarily loose. That is, not only a change in a certain direction,
Changes in any direction (as a plane) can be detected. In addition, by changing the size of the small area according to the type and size of the singular point to be detected, the degree of the singular point, the color, etc., it is possible to respond to various uses and fields that are difficult to obtain with conventional differentiation processing, Outgoing performance can be improved.

According to the present invention (claim 2), the error between the maximum value and the minimum value is compared with a predetermined threshold value for each small area, and a bitmap is created with a specific numerical value or symbol. The position, size, shape, degree, etc. of the singular point can be determined by visually observing the created bitmap. In addition, since the surface is binarized or multi-valued according to the degree of change, the shape and change state of the defect can be grasped more accurately than ever, and as a result, a unique portion such as a defect can be extracted with high sensitivity and reproducibility.

Further, according to the present invention (claim 3), since the fixed noise removing means is provided between the storage means and the maximum / minimum value detecting means, it is present in various information and sensors originally included in the inspection master. It is possible to detect only the singular point existing in the test object in a form in which the various information is removed as fixed noise without being influenced by the inherent variation or the like. In addition, an inspection in the case where the nature of the background of the surface to be inspected fluctuates becomes possible.

Further, according to the present invention (claim 4), the random noise elimination means is provided between the storage means or the fixed noise elimination means and the maximum / minimum value detection means. Random noise can be removed, and only singular points present in the test object can be detected thereafter.

Further, according to the present invention (claim 5), since a plurality of storage circuits are configured so as to simultaneously perform the storage and the arithmetic processing while changing the role alternately between the storage and the arithmetic processing. Can easily cope with continuous work such as Further, the processing speed can be increased.

Further, according to the present invention (claim 6), since the sensor is constituted by an optical camera, a singular point existing on the outer surface or inner surface of the test object is detected by reflected light or transmitted light from illumination. It is possible.

Further, according to the present invention (claim 7), since the distance between the illumination and the object to be inspected is clarified, the singular point can be detected with high accuracy.

Further, according to the present invention (claim 8), since a grating is provided between the illumination and the object to be inspected, a bright portion,
The contrast of dark areas can be clearly seen,
It is possible to improve the accuracy of detecting a singular point.

Further, according to the present invention (claim 9), the threshold value used for the determination means or the bitmap creation means is subjected to lens-specific correction, so that there is no restriction by the optical camera. Singular points can be detected.

Further, according to the present invention (claim 10), based on the bitmap created by the bitmap creating means,
Data is analyzed for a plurality of items, and the data is compared with a reference value determined in advance for each type of defect to determine the type of defect. Therefore, the type of defect and the position and degree of the defect are calculated using a computer or the like. , External shape, internal change status, etc. can be automatically instantaneously determined.

Further, according to the present invention (claim 11), there is further provided a defect judging means for judging whether or not the defect is a defect based on the width of the unique portion in each of the vertical and horizontal directions. The pass / fail can be determined based on the result.

Further, according to the present invention (claim 12), a new singularity detection method can be provided by performing a specific process in a small area, without using the conventional differential processing method. That is, since a change in illuminance as a surface, not a straight line, is detected, a small change can be detected regardless of the direction of the change on the plane to be inspected. In addition, when detecting a change in brightness on the surface to be inspected, it is possible to detect various degrees of illuminance change by changing the unit of the size of the surface (small area) for calculating the change in various ways.

[0052]

[Brief description of the drawings]

FIG. 1 is a configuration diagram showing a first embodiment of the present invention.

FIG. 2 A part of a storage circuit (raw image data near an unclear unique part)

FIG. 3 shows image data after fixed noise removal processing with respect to FIG. 2;

FIG. 4 shows image data after random noise elimination processing with respect to FIG.

5 is image data after calculating a difference between a maximum value and a minimum value for each small area with respect to FIG. 4;

FIG. 6 is a bitmap created after the binarization processing for FIG. 5;

FIG. 7 is a graph of a data analysis result with respect to FIG. 6 (projection distribution processing);

FIG. 8 shows a part of a storage circuit (raw image data in the vicinity of a distinct singular part)

FIG. 9 shows image data after fixed noise removal processing with respect to FIG.

FIG. 10 shows image data after random noise elimination processing with respect to FIG.

11 shows image data after calculating the difference between the maximum value and the minimum value for each small area with respect to FIG.

FIG. 12 is a bitmap created after the binarization processing for FIG. 11;

FIG. 13 is a graph of a data analysis result for FIG. 12 (projection distribution processing);

FIG. 14 is a configuration diagram showing a second embodiment of the present invention.

FIG. 15 is an example of a lattice

FIG. 16 shows image data after fixed noise removal processing with respect to FIG. 2;

FIG. 17 shows image data after random noise elimination processing with respect to FIG.

FIG. 18 shows image data after the conventional differentiation processing with respect to FIG.

FIG. 19 is a binarization process for FIG. 18 (threshold value 5);
Bitmap created later

FIG. 20 shows a binarization process (threshold value 1) for FIG. 18;
Bitmap created later

FIG. 21 shows image data after fixed noise removal processing with respect to FIG.

FIG. 22 shows image data after random noise elimination processing for FIG. 21;

FIG. 23 shows image data after the conventional differentiation processing with respect to FIG. 22;

FIG. 24 is a binarization process for FIG. 23 (threshold value 5);
Bitmap created later

FIG. 25 shows an example of a unique portion of a sheet surface.

FIG. 26 shows a change in illuminance on line A1-A2 in FIG.

27 shows a change in illuminance on line B1-B2 in FIG. 25;

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Sheet 2 Take-up roll 3 Lighting 4 CCD camera 4 5 Lattice 6 Slit 10 Clear unique part 11 Unclear unique part

 ──────────────────────────────────────────────────続 き Continued on the front page (72) Inventor Tohru Uemura 10 Goi Kaigan, Ichihara-shi, Chiba Asahi Glass Co., Ltd. Chiba Plant (72) Inventor Yoshiki Tanaka 38, Yawata Kaigan-dori, Ichihara-shi, Chiba Asahi Glass Engineering Co. 72) Inventor Kazuhiro Fujimoto 38 Asahi Glass Engineering Co., Ltd. 38 Yawata Kaigan-dori, Ichihara-shi, Chiba (72) Inventor Teruo Kataoka 38 Asahi Glass Engineering Co., Ltd. 38 Yawata-kaigan-dori, Ichihara-shi, Chiba

Claims (12)

[Claims] 1. A sensor for converting predetermined information included in an object to be inspected into image information, a storage unit for storing image information converted by the sensor, and the image information stored in a small area having a predetermined number of pixels each in the vertical and horizontal directions. The image in each small area is divided into a plurality of small areas shifted one pixel at a time in the vertical direction and / or the horizontal direction, and the size of the image information for each pixel is sequentially compared for each of the small areas. Maximum / minimum value detecting means for detecting the maximum value and the minimum value of the information, error calculating means for calculating an error between the maximum value and the minimum value obtained by the maximum / minimum value detecting means, and an error calculated by the error calculating means A singularity detecting device, comprising: a singularity detecting device, which compares a threshold value with a predetermined threshold value and determines that a singularity exists in the test object when the threshold value is exceeded in at least one or more small areas. . 2. The method according to claim 1, wherein the error calculated by the error calculating means is compared with a predetermined threshold value instead of the determining means.
Bitmap creation for creating a bitmap in which a specific numerical value or symbol is assigned to a pixel at a predetermined position in each of the small areas if a specific numerical value or symbol is less than the threshold value when the specific numerical value or symbol is less than the threshold value 2. The singularity detecting device according to claim 1, further comprising means. 3. The first storage circuit for storing at least image information of an inspection master having no singularity previously converted by the sensor, and an image of the inspection object converted again by the sensor. A fixed noise storing means for storing information, wherein the image information of each pixel of the first storage circuit is subtracted from the image information of each pixel of the second storage circuit, and the remaining noise is stored as image information. 3. The singularity detecting device according to claim 1, wherein a removing unit is provided between the storing unit and the maximum / minimum value detecting unit. 4. The image information stored in said storage means or the image information processed by said fixed noise removing means is converted into a small area having a predetermined number of pixels in each of the vertical and horizontal directions, one pixel at a time in the vertical and / or horizontal direction. It is divided into a plurality of shifted small areas, and the sum of the image information of each pixel is obtained for each of the small areas, divided by the number of pixels constituting the small area, and the integer part is set at a predetermined position in the small area. 2. The apparatus according to claim 1, further comprising a random noise eliminator assigned to pixels between the storage unit or the fixed noise eliminator and the maximum / minimum value detector.
4. The singular point detection device according to 2 or 3. 5. The storage means is divided into a plurality of storage circuits,
In parallel with the process of storing the image information converted by the sensor in one storage circuit, one of the other storage circuits includes the maximum / minimum value detection unit, the error calculation unit, the determination unit, and the bit map. After performing at least one of the calculation processing and the storage processing in the storage circuit, the one storage circuit performs the calculation processing on the image information stored in the one storage circuit. Value detection means, the error calculation means, the determination means,
A plurality of storages are sequentially performed, such as performing each operation of the bitmap creation unit and the random noise removal unit and performing a process of storing image information converted by the sensor in one of the other storage circuits. 5. The method according to claim 1, wherein the storage and the arithmetic processing are simultaneously performed while changing the roles of the circuit and the storage processing alternately.
The singular point detection device according to the above. 6. The sensor according to claim 1, wherein the sensor is at least one or more optical cameras that convert predetermined information included in reflected light or transmitted light from illumination to the inspection object into image information. The singular point detection device according to 3, 4, or 5. 7. The sensor receives specularly reflected light from the illumination of the inspection object, and the distance between the illumination and the inspection object is at least 1/5 of the distance between the sensor and the inspection object and 3
7. The singularity detecting device according to claim 6, wherein the singularity detecting device is separated by at least 00 mm. 8. The singularity detecting apparatus according to claim 6, further comprising a grid provided with a plurality of slits between the illumination and the object to be inspected. 9. The threshold value of the determination means or the bitmap creation means is determined by the aberration of the lens built in the optical camera and the variation inherent in each lens when a plurality of optical cameras are used. 9. A method according to claim 6, wherein a different threshold value is set for each pixel in at least one row in a predetermined direction or a different threshold value for each of a plurality of pixels in accordance with at least one of the predetermined directions.
The singular point detection device according to the above. 10. A counting means for counting the number of specific numerical values or symbols for each row or each column in at least one of a vertical direction and a horizontal direction with respect to the bitmap created by the bitmap creating means, A projection distribution creating unit that graphs the aggregation result of the aggregation unit with the aggregation number as the height direction, and at least the width of the unique portion in the vertical and horizontal directions, the width in the vertical direction and the width in the horizontal direction based on at least the projection distribution creating unit. Ratio, the area of the singular part graphed, the ratio of the area divided by the vertical width or the horizontal width, the number of irregularities in the height direction of the singular part graphed, the fixed noise removal means or the At least one item among items including a determination as to which of the plurality of predetermined setting ranges the maximum value of the image information of each pixel after each arithmetic processing by the random noise removing unit belongs to Is compared with a reference value determined in advance for each type of defect, and the defect having the highest degree of coincidence or the defect having the highest degree of coincidence after performing weighting processing for each item is determined to be the defect type. 2. The method according to claim 2, further comprising:
The singular point detection device according to 6, 7, 8 or 9. 11. When the defect type is determined by the defect type determining means, the width of each unique portion in the vertical and horizontal directions based on the projection distribution creating means is compared with a pass / fail determination width determined in advance for each defect type. 11. The singularity detection device according to claim 10, further comprising a defect determination unit that determines a defect when at least one of the direction and the lateral direction is equal to or larger than the pass / fail determination width. 12. An optical camera converts predetermined information of reflected light or transmitted light from illumination to an object to be inspected into image information, and stores the image information in a bitmap form for each pixel in a storage unit. The image information is divided into a plurality of small areas each having a predetermined number of pixels in the vertical and horizontal directions and shifted one pixel at a time in the vertical direction and / or the horizontal direction. The maximum value and the minimum value of the image information in each small area are detected by successively comparing the sizes, an error between the maximum value and the minimum value is calculated, and the calculated error is compared with a predetermined threshold value. A singularity detection method for determining that a singularity exists in the test object when a threshold value is exceeded in at least one or more small areas.