JPH09318555A - Visual inspection method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To enhance the inspection accuracy while simplifying the setting of an inspection area. SOLUTION: The image of an object 15 being picked up by means of a camera 11 is processed by means of an inspection unit 13 in order to check the external view of the object 15. When an inspection area 21 is set, a conventional method for inspecting the area 21 using an entire dictionary therefor is not employed but a dictionary is prepared for each of a plurality of small areas 31-34, 41-44 in order to enhance the inspection accuracy.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a visual inspection method for image-processing an inspection object and performing a quality inspection of the inspection object using a pattern matching method.

[0002]

2. Description of the Related Art Conventionally, as this kind of quality inspection, one dictionary is created for the entire predetermined inspection area including the inspection target image, and pattern matching is performed between this dictionary and the inspection target image (input image). There are many things. That is, the similarity r between the dictionary and the input image is calculated by, for example, the following (1)
The quality is determined based on the value obtained from the relation as in the formula. r = {NΣ (B × G) −ΣB × ΣG} / √ [{NΣB ² − (ΣB) ² } × {NΣG ² − (ΣG) ² }] ... (1) N: score of dictionary, Σ (B × G): Cross-correlation value between dictionary image and input image ΣB: Sum of brightness values of input image, ΣB ² : Autocorrelation value of input image ΣG: Sum of brightness values of dictionary image, ΣG ² : Self of dictionary image Correlation value

Further, the appearance inspection of an inspection object in which a characteristic part (a part where the brightness value changes) and a non-characteristic part (a part where the brightness value hardly changes) are mixed,
The inspection is performed manually by dividing into a characteristic region and a non-characteristic region. For example, FIG.
When there is such an image to be inspected, the areas 1 to 5 are set and inspected as shown in FIG.

[0004]

By the way, the similarity r obtained by the correlation calculation such as the above equation (1), which is used in the pattern matching, is a scratch or a stain on the area of the portion (inspection area) to be a dictionary. Since it changes depending on the ratio of the portion, if the inspection is performed by pattern matching using the entire inspection area as one dictionary as described above, the detection accuracy of scratches or stains may be insufficient. Therefore, it is conceivable to set the inspection area separately as shown by reference numerals 1 to 4 in FIG. 6B. However, since the detection accuracy of scratches, dirt, and the like changes depending on the area of the inspection area, a constant detection is performed. In order to ensure accuracy, it is necessary to specify the inspection area by dividing it so that the areas are about the same, which causes a problem of setting. Further, in many cases, the number of areas that can be designated as the inspection area is limited, and there are cases in which the setting becomes impossible depending on the required inspection accuracy.

Further, in a featureless portion (a portion in which there is almost no change in the brightness value), the similarity is accurately determined due to the influence of a minute fluctuation of the input image from the camera (due to input system noise or quantization noise). Since it cannot be calculated, there is also a problem that the inspection by pattern matching cannot be applied to a portion having no characteristic. Therefore, when inspecting an object in which a characteristic part and a non-characteristic part coexist, it is necessary to specify the inspection area separately for the characteristic part and the non-characteristic part, which is troublesome to set. In addition, depending on the object to be inspected, there are cases where the setting becomes impossible due to the limitation of the number of areas that can be designated as the inspection area. Therefore, an object of the present invention is to eliminate the deterioration of the detection accuracy for scratches and dirt and the inconvenience of setting, and to eliminate the case where setting is impossible.

[0006]

In order to solve such a problem, in the invention of claim 1, the specified inspection area is divided into rectangular small areas of arbitrary size, and a dictionary is created for each small area. By performing pattern matching between the dictionary and the image to be inspected, the detection accuracy for scratches and stains is prevented from deteriorating. At this time, by setting the size of the rectangular small area to any size, it is possible to adjust the detection accuracy for scratches and dirt. According to the second aspect of the present invention, the variance value of the brightness value of the dictionary is obtained for each small area, and the image to be inspected is inspected by the variance value of the brightness value in the area where the variance value is less than or equal to the set value. In a larger area, inspection is performed by pattern matching between the dictionary and the image to be inspected. As a result, it becomes possible to perform highly accurate inspection with simple settings even for an inspection target in which characteristic portions and non-characteristic portions are mixed.

According to the third aspect of the present invention, the division of the areas is performed so that the small areas partially overlap each other, thereby stabilizing the detection accuracy at the boundary of the small areas. Further, in the invention of claim 4, when the inspection area having an arbitrary shape is divided into small areas, the division position of the small area is corrected so that the area of the inspection image existing in the small area is maximized. The detection accuracy is stabilized when only a small portion of the arbitrarily shaped area is included in the small area.

[0008]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS FIG. 1 is an explanatory diagram for explaining an embodiment of the present invention. FIG. 3A is a schematic diagram of an image processing system in which the present invention is implemented, FIG. 2B is an inspection region image example, and FIGS.
That is, according to the present invention, the inspection target 15 is imaged by the camera 11 as an imaging device and the image signal 12 of the inspection target 15 is imaged by the system shown in FIG.
The quality is judged there and the judgment result signal 14 is outputted. Up to this point, the processing is the same as the conventional one, but the present invention further performs the following processing.

Now, after the reference image is captured, the inspection area 21 to be inspected is designated as shown in FIG.
An instruction is given to the image processing inspection device 13. The inspection device 13 replaces the designated area 21 with a rectangular small area (here, 31, 32, in this case) of a preset size as shown in FIG.
33, 34), and each small area 31, 32,
A dictionary for pattern matching is created for each of 33 and 34. After that, capture the image of the inspection target (input image),
Inspection by pattern matching is performed for each small area. Although the small areas are divided so as not to overlap with each other here, as shown in FIG. 1D, the start points (for example, upper left coordinates) of the small areas are shifted so that the small areas overlap. can do. Each small area at that time is coded 4
1, 42, 43 and 44, and the overlapping area is indicated by reference numeral 45. By doing so, it is possible to absorb fluctuations in the vicinity of the boundary of each small region and stabilize the detection accuracy.

FIG. 2 is a flow chart showing another embodiment of the present invention, and FIG. 3 is an explanatory diagram for specifically explaining FIG. First, after the reference image is captured (see step S1 in FIG. 2), the inspection area 51 to be inspected is designated as shown in FIG. 3A and the image processing inspection device 13 is instructed. The inspection device 13 divides the designated area 51 into rectangular small areas 61 and 62 (61: small dispersion value, 62: large dispersion value) of a preset size, for example, as shown in FIG. Subsequently, here, the variance value of the brightness values is calculated for each of the divided small areas. Further, the calculated dispersion value is stored, and in the area where the dispersion value is large, a dictionary for pattern matching is created as in the case of FIG.

At the time of executing the subsequent inspection, the image to be inspected (input image) is taken in, and the inspection by pattern matching is performed in the region where the variance value of the reference image is larger than the set value (steps S2 to S8 and S12 in FIG. 2). In a region in which the variance value of the reference image is less than or equal to the set value, the inspection is performed by comparing the variance value with the corresponding region of the inspection target image (steps S2, S3, S7 to S1 in FIG. 2).
2). In the above, the inspection is performed by comparing the variance values. However, as long as it is an effective means for inspecting a featureless portion, for example, the inspection is performed by comparing the maximum value and the minimum value of the luminance value, or the luminance in a certain range. The inspection may be performed by the number of pixels having a value. Also, small area 6
It goes without saying that the division of 1, 62 can be performed so that a part thereof overlaps as shown in FIG.

FIG. 4 is a flow chart showing still another embodiment of the present invention, and FIG. 5 is an explanatory view for specifically explaining FIG. This describes a method of dividing an inspection area having an arbitrary shape. When there is an inspection area 71 as shown in FIG. 5A, first, a circumscribed rectangle 81 of the inspection area is made as shown in FIG. 5B (see step S1 in FIG. 4).
After dividing this into Y-sized strips (step S in FIG. 4).
2), the following processing is executed for each band.

That is, a search is performed from the left end of the band to find the left end of the inspection area, and its X coordinate position is set to the left end X0 of the first small area, and a small area whose left end is X0 is created (step S3 in FIG. 4). (See S4). The size of the small area is set in advance according to the inspection target.
Next, a search is performed from the left end of the small area to find the left end X1 of the inspection area within the small area (see step S5 in FIG. 4). If the found X1 is the same as the left edge X0 of the small area, the search is performed from the right edge of the small area to find the right edge X2 of the inspection area within the small area (see steps S6 and S7 in FIG. 4). When X2 is found, the division position of the small area is corrected so that the right end of the small area is set to X2 (step S in FIG. 4).
8), if not, the division position of the small area is corrected so that the left end of the small area is set to X1 (step S in FIG. 4).
7 ').

Thereafter, a search is performed from the upper end of the small area to find the upper end Y1 of the inspection area within the small area (see step S9 in FIG. 4). When the found Y1 is the same as the upper end of the small area, a search is performed from the lower end of the small area to find the lower end Y2 of the inspection area within the small area (step S10 in FIG. 4).
See S11). If Y2 is found, the bottom of the small area is set to Y
The division position of the small area is corrected so as to be 2 (see FIG.
(See step S12), otherwise, the division position of the small area is corrected so that the upper end of the small area is Y1 (see step S11 'in FIG. 4). Subsequently, the small area having the right end coordinates of the small area after position correction as the left end is set as the next area, and the same position correction as described above is performed. After that, when it hits the right end of the band (see step S13 in FIG. 4), it moves to the next band and continues the division processing (steps S14, S15 in FIG. 4).
(See S16).

[0015]

According to the first aspect of the present invention, the specified inspection area is divided into rectangular small areas of arbitrary size, a dictionary is created for each small area, and a pattern is created between the dictionary and the image to be inspected. By performing matching, it is possible to prevent deterioration of detection accuracy for scratches and dirt. At this time, if the size of the rectangular small area is arbitrary, the detection accuracy for scratches and dirt can be adjusted. According to the second aspect of the present invention, the variance value of the brightness value of the dictionary is obtained for each small area, and the image to be inspected is inspected by the variance value of the brightness value in the area where the variance value is less than or equal to the set value. Since inspection is performed by pattern matching between the dictionary and the image to be inspected in a larger area, high precision inspection can be performed with simple settings even for inspection objects in which characteristic parts and non-characteristic parts coexist. It will be possible.

According to the third aspect of the present invention, by dividing the areas so that the small areas partially overlap each other, it is possible to stabilize the detection accuracy at the boundaries of the small areas. Further, in the invention of claim 4, when the inspection area having an arbitrary shape is divided into small areas, the division position of the small areas is corrected so that the area of the inspection image existing in the small areas becomes maximum. Therefore, it is possible to stabilize the detection accuracy in the case where only a small portion of the arbitrarily shaped area is included in the small area.

[Brief description of drawings]

FIG. 1 is an explanatory diagram for explaining an embodiment of the present invention.

FIG. 2 is a flowchart showing another embodiment of the present invention.

FIG. 3 is an explanatory diagram for specifically explaining FIG. 2;

FIG. 4 is a flow chart showing still another embodiment of the present invention.

FIG. 5 is an explanatory diagram for specifically explaining FIG. 4;

FIG. 6 is a schematic explanatory diagram of a conventional method.

[Explanation of symbols]

1, 2, 3, 4, 5, 21, 51, 71 ... Inspection area, 1
1 ... Camera, 12 ... Image signal, 13 ... Image processing inspection device, 14 ... Judgment result signal, 15 ... Inspection object, 31, 3
2, 33, 34, 41, 42, 43, 44, 82 ... Small area, 45 ... Overlapping area, 61 ... Small area (small variance value), 62
... Small area (large variance value), 81 ... Circumscribing rectangle.

Claims (4)

[Claims] 1. When performing image processing on an inspection target to inspect its quality, a predetermined inspection region including a reference inspection target image is divided into rectangular small regions of arbitrary size, and each divided small region is divided. A visual inspection method is characterized in that a dictionary for pattern matching is created, and the inspection is performed by pattern matching between the dictionary and the image to be inspected for each small area. 2. The dispersion value of the brightness value of the dictionary is obtained for each of the small areas, and in the area where the calculated dispersion value is less than or equal to the set value, inspection is performed based on the dispersion value of the brightness value, and the dispersion value is set. 2. An area larger than the value is inspected by pattern matching between a dictionary and an image to be inspected.
Appearance inspection method described in. 3. The small area is divided so that a part thereof overlaps each other.
The visual inspection method according to any one of 1. 4. When dividing the predetermined inspection area into small areas, the division position of the small area is corrected so that the area of the inspection area in the small area is maximized. The visual inspection method according to any one of 1.