JP2004318488A - Product inspection method and product inspection device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To solve the problem that a conventional product inspection which detects the incorporation of the different kind of product and a defective product by performing pattern matching using a registered master pattern requires time and man-hours for registering the master pattern, resulting in high inspection cost. <P>SOLUTION: By performing pattern matching mutually between products included in the aggregate of products, the processes of production and registration of the master pattern are eliminated. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a product inspection method and a product inspection device for performing a product inspection using an image processing technique, and in particular, mixing of different types of products in a product package in which a plurality of products are integrated into one package, and the quality of product packaging. The present invention relates to a product inspection method and an inspection device for inspecting a product.
[0002]
[Prior art]
In a product inspection using an image processing technique, a product inspection method of matching a product pattern using a template including a registered master pattern is used. For example, in Patent Document 1, a pattern using a master pattern is used. When an inspection is performed by matching, a comparison method for comparing a threshold image with a difference image between an inspection target image and a master pattern is improved to prevent erroneous recognition due to a variation in the size of the inspection target.
[0003]
[Patent Document 1]
JP 2002-183733 A
[Problems to be solved by the invention]
In a product inspection method using a master pattern as in the method of Patent Document 1, it is necessary to register the master pattern in advance, so that there is a problem that registration of the master pattern requires time and man-hours.
[0005]
There are several to several tens of patterns to be inspected for one product, and in order to inspect a wide variety of products, it is necessary to register a considerable number of master patterns in advance, and the work man-hour is burdensome.
[0006]
An object of the present invention is to reduce the time and labor required for pre-registration of a master pattern and to reduce manufacturing costs.
[0007]
Even when the inspection is performed by pattern matching using a registered master pattern, it is desirable to reduce the number of steps for creating the master pattern as much as possible.
[0008]
Another object of the present invention is to reduce inspection costs and, consequently, manufacturing costs by reducing the number of steps for creating a master pattern.
[0009]
Another object of the present invention is to improve inspection accuracy and reliability in a technology for inspecting a product by pattern matching.
[0010]
[Means for Solving the Problems]
The above object of the present invention is achieved by the following invention.
[0011]
1. In a product inspection method of photographing an aggregate of a plurality of products, converting the image data into image data, processing the image data, and inspecting whether or not a heterogeneous product or a defective product is mixed in the aggregate.
A product inspection method, wherein inspection is performed by pattern matching for comparing the image data of each of the plurality of products with each other.
[0012]
2. A plurality of the aggregates are continuously supplied to an inspection station, and when an inspection is performed, the image data acquired from the aggregates in a previous inspection is used for an inspection of the subsequent aggregate. 2. The product inspection method according to the above 1.
[0013]
3. Photographing an aggregate of a plurality of products, converting the image data into image data, performing pattern matching between the image data and a registered reference pattern, and checking whether the aggregate contains a heterogeneous product or a defective product. In the product inspection method of inspecting whether or not
A product inspection method, wherein one area to which the reference pattern is applied is determined, and the position of another area to which the reference pattern is applied is determined by coordinate calculation based on the one area.
[0014]
4. 4. The product inspection method according to the item 3, wherein the position of the other area determined by the coordinate calculation is corrected by pattern matching using the reference pattern.
[0015]
5. In a product inspection method of photographing an aggregate of a plurality of products, converting the image data into image data, processing the image data, and inspecting whether or not a heterogeneous product or a defective product is mixed in the aggregate.
Using the reference pattern, the entire inspection area is scanned and pattern matching is performed, an overall model correlation value for the reference pattern is calculated, and pattern matching is performed using the divided image obtained by dividing the reference pattern. A product inspection method comprising: calculating a divided model correlation value obtained as described above; and performing an inspection using the overall model correlation value and the divided model correlation value.
[0016]
6. In the product inspection method of inspecting the mixture of different products or defective products by pattern matching,
At least using the correlation value obtained by pattern matching using the reference pattern, the normalized correlation value obtained by pattern matching, and the area difference factor, determining the similarity between the reference pattern and the target product. Product inspection method.
[0017]
7. A product inspection method for inspecting the mixture of different products or defective products by pattern matching, wherein the product inspection is performed based on two-dimensional image data and one-dimensional image data of a projected image projected from one direction. Method.
[0018]
8. In a product inspection apparatus having an imaging unit that generates image data by imaging and an image processing unit that performs image processing on the image data generated by the imaging unit,
A product inspection apparatus, wherein the image processing means detects mixing of different types of products or mixing of defective products in a set of products by pattern matching for comparing the image data of each of the plurality of products with each other.
[0019]
9. The image processing means, when performing an inspection of an aggregate of products supplied continuously, using the image data acquired from the aggregate in a previous inspection for an inspection of the subsequent aggregate. 9. The product inspection device according to the above item 8.
[0020]
10. In a product inspection apparatus having an imaging unit that generates image data by imaging and an image processing unit that performs image processing on the image data generated by the imaging unit,
The image processing means determines one area to which the reference pattern is applied, and determines the position of another area to which the reference pattern is applied by coordinate calculation based on the one area. And product inspection equipment.
[0021]
11. The product inspection apparatus according to claim 10, wherein the image processing unit corrects the position of the other area determined by the coordinate calculation by pattern matching using the reference pattern.
[0022]
12. In a product inspection apparatus having an imaging unit that generates image data by imaging and an image processing unit that performs image processing on the image data generated by the imaging unit,
The image processing means performs pattern matching by scanning the entire inspection area using a reference pattern, calculates an overall model correlation value for the reference pattern, and generates a divided image obtained by dividing the reference pattern. A product inspection apparatus comprising: calculating a divided model correlation value obtained by performing pattern matching using the same; and performing an inspection using the overall model correlation value and the divided model correlation value.
[0023]
13. In a product inspection apparatus having an imaging unit that generates image data by imaging and an image processing unit that performs image processing on the image data generated by the imaging unit,
The image processing means determines a similarity between the reference pattern and the target product using at least a correlation value obtained by pattern matching using the reference pattern, a normalized correlation value obtained by pattern matching, and an area difference factor. A product inspection device.
[0024]
14． A product that has image capturing means for generating image data by image capturing and image processing means for performing image processing on the image data generated by the image capturing means, wherein the image processing means performs product inspection by pattern matching using two-dimensional image data In the inspection device,
The product inspection apparatus, wherein the image processing means converts the two-dimensional image data into one-dimensional image data of a projected image projected from one direction, and performs a product inspection based on the one-dimensional image data.
[0025]
BEST MODE FOR CARRYING OUT THE INVENTION
FIG. 1 schematically shows a product inspection apparatus according to an embodiment of the present invention.
[0026]
Reference numeral 1 denotes a package as an aggregate of products to be inspected, for example, a package in which twenty films of the same type are combined into one. The market is supplied in the form of a package 1. Accordingly, transportation and handling in the process from shipment at the factory to delivery to the retail store are performed in the form of the package 1. In the following description, the package 1 to be inspected is called a work.
[0027]
Reference numeral 2 denotes a photographing unit including a two-dimensional CCD camera, which photographs the work 1 and generates image data. FIG. 2 shows an example of an image photographed by the photographing means 2.
[0028]
Reference numeral 3 denotes an image processing unit including a personal computer or the like, which processes image data generated by the photographing unit 1. The contents of the processing are mainly image processing such as contour enhancement, pattern recognition and pattern matching, as described below.
<Inspection without using registered master pattern>
The present invention provides a method and an apparatus for performing an inspection using image data obtained by photographing a work without using a registered master pattern.
[0029]
Such a product inspection method will be described with reference to FIG.
An image W1 of a workpiece is obtained from image data obtained by photographing, a partial photograph of one product is cut out from the image W1, and this is used as a template T for pattern matching. The partial photograph is cut out by performing image processing on the image W1, recognizing the repetitive pattern, and determining the outline of the rectangle forming the template T from the recognized repetitive pattern.
[0030]
In the following description, one product and one product forming the patterns P1 to Pn are referred to as individual products.
[0031]
For example, the position of the first pattern P1 is set as a reference position, the pattern P1 is set as a reference pattern, that is, a template T, and pattern matching is performed with the pattern P2. Thereafter, by pattern-matching the pattern Pm and the pattern Pm + 1, an inspection is performed on all the individual products of one work.
[0032]
FIG. 3 is a flowchart of an inspection process in the product inspection method according to the embodiment of the present invention.
[0033]
In step S1, a work is photographed to acquire image data.
In step S2, preprocessing is performed on the acquired image data. The preprocessing is a process for enhancing a pattern included in an image, such as edge enhancement, to accurately and easily extract an inspection window and a divided image described below.
[0034]
In step S3, an inspection window is cut out. FIG. 4 is a conceptual diagram of the inspection window cutout. As shown, ten rectangular inspection windows W2 are set from the contours of ten individual products existing in the image of the edge-emphasized work. The inspection window W2 corresponds to the patterns P1 to Pn in FIG.
[0035]
In step S4, the inside of the window W2 at the predetermined reference position is divided into, for example, a plurality of divided images DW1, DW2, and DW3 as shown in FIG. For example, as shown in FIG. 5, the divided images DW1, DW2, and DW3 are regions that specify a characteristic pattern existing in the window, and the divided images DW1, DW2, and DW3 are patterns that represent the characteristic pattern. It is created by setting a rectangular area that is specified by recognition and has a predetermined amount of a small margin added around the specified characteristic pattern.
[0036]
In step S5, the entire image and the divided image of the individual product at the reference position, for example, the position of the pattern P1 in FIG. Perform pattern matching between them. The entire image is, for example, an image of the entire inspection window as indicated by a template T in FIG. Such pattern matching is repeatedly performed on the patterns Pm and Pm + 1, and pattern matching is performed on all individual products of one work.
[0037]
In step S6, a whole model correlation value obtained by pattern matching of the whole image and a divided model correlation value obtained by pattern matching of the divided image are calculated.
[0038]
Steps S5 and S6 are performed for all individual products in one work, and data of one work is obtained.
[0039]
In step S7, the pass / fail of the work is determined by comparing the data of one work including the calculated overall model correlation value and the divided model correlation value with a threshold value.
[0040]
An example of the pass / fail judgment algorithm in step 7 will be described in the case of a photographic film product.
[0041]
Inspection by pattern matching is basically performed by determining the magnitude of a correlation value obtained by pattern matching. Correlation values are high between identical or similar patterns, and low between different patterns. Therefore, the correlation value of the same product is high, but the correlation value of the different product or defective product is low, so that the different product or the defective product can be detected by the determination using the threshold. The image shown in FIG. 6 is used as a template for pattern matching.
[0042]
The entire template (whole image) in FIG. 6A is used as a base material for determining the pass / fail of each individual product, that is, the pass / fail of the product. In the pattern matching using the entire template, the horizontal and vertical characters are packed vertically in different directions so that the characters are written in a horizontal direction. In particular, when they are mixed, a large difference appears in the correlation value, so that the cause of these defects can be reliably detected.
[0043]
The product name template (divided image) in FIG. 6B is effective for detecting mixing of different types.
[0044]
The sensitivity template (divided image) shown in FIG. 6C is effective for detecting mixing of film products having different sensitivities.
[0045]
The number template (divided image) in FIG. 6D is effective for detecting the mixture of film products having different numbers.
[0046]
By pattern matching using the entire surface template of FIG. 6A, it is possible to identify most of various types of film products, and it is possible to detect mixing of different types of products. Regarding different types of products of the same type of film or different types of products with a specific similar design, pattern matching using the entire template shows a high correlation value for both, making it impossible to reliably detect contamination of different types of products There are cases.
[0047]
In this case, pattern matching using the product name template, the sensitivity template, and the number template is performed in an appropriate combination.
[0048]
Even when there is no discernible difference in the correlation value of the pattern matching by the whole-surface template, any of the correlation values of the pattern matching of the product name template, the sensitivity template, and the number template has a discernable difference, so that a different type of product or defective Product contamination can be reliably detected.
[0049]
In addition, individual products that are packed vertically or horizontally or that are packed upside down can be reliably detected by pattern matching using the various templates described above.
[0050]
The data of individual products are integrated, and if there is no single failed individual product, it is determined to be a passed work, and the others are determined to be failed work.
[0051]
If it passes (Yes in S8), in step S9, the image data of the entire image and the divided image at the reference position is stored in the memory of the image processing means as data of the reference pattern. The stored data is used in pattern matching for the next individual product. Data of failed individual products is not stored. Therefore, in the inspection of the work next to the work determined to be rejected, the data created in the inspection before the inspection of the work determined to be rejected is used.
[0052]
When the inspection of one work is completed, steps S1A to S4A equivalent to steps S1 to S4 are performed for the next work, and in step S10 following step S4A, the data stored in the memory in the previous inspection is used as a reference. Perform pattern matching as a pattern.
[0053]
Subsequent to step S10, steps S6A to S9A equivalent to steps S6 to S9 are performed, and the inspection of the next work is completed.
[0054]
Thereafter, by repeating steps S1A to S9A, the continuously supplied workpieces are inspected, and when the inspection of all the workpieces is completed (Yes in S11), the process ends.
<Inspection using registered master pattern>
In the case where the probability of mixing different types in one package is high, using the method described above, the master pattern is not updated for each work, but is carried over to the next next work. More often. In addition, when the reference position is set, the individual products present at the reference position are rejected, and the setting process of the master pattern becomes complicated. is there.
[0055]
Therefore, when there is a high probability of mixing different types, a conventional method of registering a master pattern in advance is used. Even when a registered master pattern is used, it is desirable to reduce the number of steps of master pattern registration. In the present invention, a master pattern including a plurality of patterns is created using coordinate calculation.
[0056]
As shown in FIG. 7A, a template including a pattern P1 is created. The pattern P1 is obtained by photographing a product using the two-dimensional CCD camera 2 in FIG. 1 and processing an image acquired by photographing. It was acquired and registered.
[0057]
In the case of a work including 10 individual products, the positions p1 to p10 from the patterns P2 to P10 are determined by coordinate calculation, and the pattern P1 is attached to the positions p1 to p10, as shown in FIG. Reference patterns P1 to P10 are created. The coordinate calculation may be a simple one consisting of multiplication on the pattern P1, but it is also possible to correct the positions of the patterns P2 to P10 as follows.
[0058]
The work is photographed by the photographing means 2 and the acquired image data is subjected to image processing to acquire image data.
[0059]
The acquired image data is subjected to pattern matching using the reference pattern P1 shown in FIG. 7A, and the maximum value of the correlation value is extracted to obtain the positions of the respective reference patterns P1 to P10. The position determined by the coordinate calculation is corrected based on the position of each pattern determined in this way. For example, correction such as replacing a part of the position obtained by the coordinate calculation with the position obtained by the pattern matching is performed.
[0060]
Pattern matching is performed on the corrected inspection area using the reference pattern. By using the correction calculation for correcting the position obtained from the coordinate calculation, the position accuracy can be improved and the probability of erroneous detection can be reduced. .
<Improvement of detection accuracy and detection reliability>
The invention described below can be applied to both a product inspection method using a registered master pattern and a product inspection method not using a registered master pattern.
[0061]
The inspection by pattern matching using the template of the whole image and the template of the divided image, which is performed in step 5 of the inspection process shown in FIG. 3, can be used for the inspection using the registered master pattern.
[0062]
That is, a template of a master pattern registered in advance is prepared, and the position of an individual product is detected by scanning an image in an inspection window of the template. The inspection window is set at the detected position, the inspection window is divided, a divided image is obtained, and a template of the divided image shown in FIG. 5 is created from the divided image.
[0063]
Pattern matching is performed using the registered master pattern template and the divided image template to obtain an overall model correlation value and a divided model correlation value.
[0064]
From the obtained correlation values, as described above in the inspection for pattern-matching the patterns of the individual products with each other, it is possible to determine with a high degree of accuracy between a passing product and a rejecting product.
[0065]
As described above, mixing of different types of products and defective products can be detected with high accuracy by pattern matching using various templates. In the present invention, a normalized correlation value is used as a correlation value obtained in pattern matching. At the same time, by adding the area difference factor as an evaluation factor for evaluating the similarity of the pattern, the accuracy of the pass / fail judgment is further improved.
[0066]
The normalized correlation value is a correlation value obtained by setting an inspection window of a fixed shape having a fixed area and executing pattern matching in the inspection window. The area difference factor is the number of pixels having a density equal to or higher than a certain value in the inspection window.
[0067]
Using the correlation value obtained by pattern matching using a template of a size and shape corresponding to the pattern as shown in FIGS. 5 and 6, the normalized correlation value, and the area difference factor, the pass / fail of the product is determined. Is determined, it is possible to remove noise and the like due to the influence of illumination light when photographing the product, and the accuracy and reliability of the inspection can be further improved.
[0068]
In pattern matching using such a normalized correlation value, an inspection error may occur because the area of the inspection target area is small.
[0069]
In the present invention, such a problem is solved by using the following method.
One-dimensional data is created by projecting the two-dimensional image data onto coordinate axes in an arbitrary direction, and the created one-dimensional data is used for inspection.
[0070]
That is, as shown in FIG. 10, the data expanded to two-dimensional coordinates, that is, two-dimensional image data is projected onto the coordinate axis X, and the data expanded to one-dimensional coordinates, that is, one-dimensional image data is obtained.
[0071]
The obtained one-dimensional image data is used, for example, for weighting the correlation value.
[0072]
【The invention's effect】
According to the first, second, eighth and ninth aspects of the present invention, since there is no need to register a master pattern for pattern matching, the amount of work required for inspection is reduced, and inspection costs can be reduced.
[0073]
According to the second or ninth aspect of the present invention, since there is no need to create a template when the inspection target is switched, the inspection speed can be increased, and a highly reliable inspection can be performed.
[0074]
According to the third, fourth, tenth, and eleventh aspects of the present invention, the amount of work and time required for creating and registering a master pattern can be reduced, and inspection costs can be reduced.
[0075]
According to the fourth or eleventh aspect of the present invention, a master pattern in which the position of each product in the product assembly is accurately reflected is created. Inspection with excellent properties becomes possible.
[0076]
According to any one of the fifth to seventh and twelfth to twelfth aspects of the present invention, erroneous detection of mixing of different types of products or defective products is reduced, and an inspection with excellent reliability and stability can be performed.
[Brief description of the drawings]
FIG. 1 is a view schematically showing a product inspection apparatus according to an embodiment of the present invention.
FIG. 2 is a photograph of a work to be inspected.
FIG. 3 is a flowchart of an inspection process in the product inspection method according to the embodiment of the present invention.
FIG. 4 is a conceptual diagram of inspection window clipping.
FIG. 5 is a diagram showing a divided image.
FIG. 6 is a diagram showing various templates.
FIG. 7 is a diagram illustrating setting of a position of an area to which a reference pattern is applied;
FIG. 8 is a diagram showing a one-dimensional image obtained by projecting a two-dimensional image.
[Explanation of symbols]
1 work 2 photographing means 3 image processing means W1, W3 image T template P1, Pm, Pm + 1, Pn pattern W2 window DW1, DW2, DW3 divided image

Claims (14)

In a product inspection method of photographing an aggregate of a plurality of products, converting the image data into image data, processing the image data, and inspecting whether or not a heterogeneous product or a defective product is mixed in the aggregate.
A product inspection method, wherein inspection is performed by pattern matching for comparing the image data of each of the plurality of products with each other. A plurality of the aggregates are successively supplied to an inspection station, and when an inspection is performed, the image data acquired from the aggregates in a previous inspection is used for an inspection of the subsequent aggregate. The product inspection method according to claim 1. Photographing an aggregate of a plurality of products, converting the image data into image data, performing pattern matching between the image data and a registered reference pattern, and checking whether the aggregate contains a heterogeneous product or a defective product. In the product inspection method of inspecting whether or not
A product inspection method, wherein one area to which the reference pattern is applied is determined, and the position of another area to which the reference pattern is applied is determined by coordinate calculation based on the one area. 4. The product inspection method according to claim 3, wherein the position of the other area determined by the coordinate calculation is corrected by pattern matching using the reference pattern. In a product inspection method of photographing an aggregate of a plurality of products, converting the image data into image data, processing the image data, and inspecting whether or not a heterogeneous product or a defective product is mixed in the aggregate.
Using the reference pattern, the entire inspection area is scanned and pattern matching is performed, an overall model correlation value for the reference pattern is calculated, and pattern matching is performed using the divided image obtained by dividing the reference pattern. A product inspection method comprising: calculating a divided model correlation value obtained as described above; and performing an inspection using the overall model correlation value and the divided model correlation value. In the product inspection method of inspecting the mixture of different products or defective products by pattern matching,
At least using the correlation value obtained by pattern matching using the reference pattern, the normalized correlation value obtained by pattern matching, and the area difference factor, determining the similarity between the reference pattern and the target product. Product inspection method. A product inspection method for inspecting the mixture of different products or defective products by pattern matching, wherein the product inspection is performed based on two-dimensional image data and one-dimensional image data of a projected image projected from one direction. Method. In a product inspection apparatus having an imaging unit that generates image data by imaging and an image processing unit that performs image processing on the image data generated by the imaging unit,
A product inspection apparatus, wherein the image processing means detects mixing of different types of products or mixing of defective products in a set of products by pattern matching for comparing the image data of each of the plurality of products with each other. The image processing unit is characterized in that, when performing an inspection of an aggregate of products supplied continuously, the image data acquired from the aggregate in a previous inspection is used for an inspection of the subsequent aggregate. The product inspection apparatus according to claim 8, wherein In a product inspection apparatus having an imaging unit that generates image data by imaging and an image processing unit that performs image processing on the image data generated by the imaging unit,
The image processing means determines one area to which the reference pattern is applied, and determines the position of another area to which the reference pattern is applied by coordinate calculation based on the one area. And product inspection equipment. The product inspection apparatus according to claim 10, wherein the image processing unit corrects the position of the other area determined by the coordinate calculation by pattern matching using the reference pattern. In a product inspection apparatus having an imaging unit that generates image data by imaging and an image processing unit that performs image processing on the image data generated by the imaging unit,
The image processing means performs pattern matching by scanning the entire inspection area using a reference pattern, calculates an overall model correlation value for the reference pattern, and generates a divided image obtained by dividing the reference pattern. A product inspection apparatus comprising: calculating a divided model correlation value obtained by performing pattern matching using the same; and performing an inspection using the overall model correlation value and the divided model correlation value. In a product inspection apparatus having an imaging unit that generates image data by imaging and an image processing unit that performs image processing on the image data generated by the imaging unit,
The image processing means determines a similarity between the reference pattern and the target product using at least a correlation value obtained by pattern matching using the reference pattern, a normalized correlation value obtained by pattern matching, and an area difference factor. A product inspection device. A product that has image capturing means for generating image data by image capturing and image processing means for performing image processing on the image data generated by the image capturing means, wherein the image processing means performs product inspection by pattern matching using two-dimensional image data In the inspection device,
The product inspection apparatus, wherein the image processing means converts the two-dimensional image data into one-dimensional image data of a projected image projected from one direction, and performs a product inspection based on the one-dimensional image data.

Images