JPH0894335A - Image processing detector - Google Patents

Abstract

PURPOSE: To identify an inspection object with a high recognition property and in a high speed by extracting a characteristic massive figure of the inspection object from image data which are picked up and creating a histogram of the amount of characteristics by measuring the amount of characteristics of each massive figure. CONSTITUTION: Image data picked up by an image pick-up camera 10 are taken into an image memory 14 by a take-in circuit 12, the characteristic massive figure of the image data is extracted by an extraction circuit 16, and the area of the extracted massive figure is measured by an area measuring circuit 18. An area histogram pattern creation means 20 has inspection objects including three massive figures with a size exceeding a certain level, for example circle, triangle, and square and the area values are 560, 540, and 750, respectively. No uniform class width is provided as the area histogram of the inspection object and a width of 500-600 and that of 700-800 are set to the circle/triangle and the square, respectively. A creation means 20 creates a master pattern for all flowing objects and identify 20 the targets with the pattern as a mask.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing inspection device for inspecting whether a known object having a characteristic pattern has been correctly conveyed or not, during conveyance.

[0002]

2. Description of the Related Art When an object is known in advance and its design is characteristic, pattern matching is conventionally performed when inspecting whether or not a planned object has been correctly conveyed. Ordinarily, geometrical characteristic quantities such as area and perimeter are measured, and the measured values are used for inspection.

Also, in the case of using the histogram of the area, the similarity is compared with those of each standard predetermined in the grade, as in Japanese Patent Laid-Open No. 5-46731, by obtaining the frequency distribution of the area and standardizing it. It was something to do.

[0004]

However, in the inspection of the physical distribution system, high-speed processing is required, so that pattern matching is unsuitable when the object is affected by rotation or displacement.

Further, even when the geometrical feature amount is used, the feature amount may not be stably obtained due to the influence of the rotation and displacement of the object.

Therefore, Japanese Patent Laid-Open No. 5-46731 discloses a method that uses an area which is a characteristic amount that is less affected by the rotation, displacement, etc. of an object. This method is used to inspect an object to be inspected. In that case, since a histogram having a uniform width with respect to the measured area value is created, it is not stable, and it is expected that processing speed will be increased due to the large amount of data.

The present invention has been made in order to solve the above-mentioned conventional problems, and the width of the classes and the number of classes of the feature amount histogram pattern such as area are the minimum necessary for distinguishing inspection objects from each other. It is an object of the present invention to provide an image processing inspection device which can be set to, and which has a high recognition accuracy and high speed with a simple method.

[0008]

SUMMARY OF THE INVENTION The present invention is an image processing inspection device for inspecting a known object having a characteristic pattern during transportation, and an image pickup means for taking an image of the object to be inspected, and an image pickup means. An image memory in which the extracted image data is captured, an extraction unit for extracting a characteristic block figure included in the inspection object from the image data in the image memory, and a characteristic amount of each of the extracted block figures is measured. The feature amount measuring means and the feature amount histogram pattern creating means for creating the feature amount histogram pattern of the inspection object by the combination of the obtained feature amount values, and the feature amount histogram pattern thus created are used to detect the inspection object. The above-mentioned object is achieved by providing an identifying means for identifying.

Further, the width and number of classes of the feature amount histogram pattern can be changed.

Further, the class width and the number of classes of the feature amount histogram pattern are corrected according to the discrimination result between the inspection object and other objects.

Further, two or more types of feature quantities of each block figure are used in combination.

[0012]

In the present invention thus constructed, the extraction means extracts a characteristic block figure of the inspection object from the image data taken by the image pickup means. The feature amount measuring means measures the feature amount (for example, area) of each of the extracted block figures. The feature amount histogram pattern creating means creates a feature amount histogram pattern used when identifying the inspection target object based on the combination of the obtained feature amount values.
The identification unit identifies the inspection object by using the created feature amount histogram pattern.

Therefore, according to the present invention, an image processing inspection apparatus having a high recognition accuracy and a high speed can be provided by a simple method by creating a feature amount histogram pattern in accordance with the characteristics of the pattern of the inspection object. It is possible to provide.

Particularly, when the class width and the number of classes of the feature amount histogram pattern can be changed, these can be set arbitrarily, so that the inspection object can be inspected by collating a certain limited block figure. You can identify things.

Further, in the case where the class width and the number of classes of the feature amount histogram pattern are corrected according to the discrimination result of the inspection object and other objects, the class width and the class The feature amount histogram pattern can be optimized by repeating the correction and verification of the feature amount histogram pattern by changing the number.

In particular, when the characteristic amount is the area of each block figure, it is unlikely to be affected by the rotation or displacement of the object. Note that the feature amount is not limited to the area, and may be any other feature amount as long as it can be numerically obtained as an image feature amount at high speed, such as a perimeter and a circularity, and two or more types. It is also possible to improve the identification accuracy of the inspection target object by using the feature amounts in combination.

[0017]

Embodiments of the present invention will be described below in detail with reference to the drawings.

In this embodiment, as shown in FIG. 1, an image pickup camera 10 for picking up an image of an object to be inspected, an image memory 14 in which image data picked up by the image pickup camera 10 is taken in through a taking-in circuit 12, From the image data on the image memory 14, an extraction circuit 16 that extracts a characteristic block figure included in the inspection object, an area measurement circuit 18 that measures the area of each extracted block figure, and An area histogram pattern creating means (software) 20 for creating an area histogram pattern of the inspection object by a combination of areas, and an identifying means (software) 22 for identifying the inspection object using the created area histogram pattern. It consists of

The operation of the embodiment will be described below.

First, the image data taken by the image pickup camera 10 is taken into the image memory 14 by the taking-in circuit 12. Next, the extraction circuit 16 extracts a characteristic block figure of the image data, and the area of each block figure is measured by the area measuring circuit 18. For example, as shown in FIG. 2, there is an inspection object 30 including three lump figures having a size larger than a certain standard, for example, a circle 32, a triangle 34, and a square 36, and the area value measured by the area measuring circuit 18 is , 560, 540, and 750, respectively. Here, as the area histogram of this inspection object, as shown in FIG.
The width of the class of 700 to 800 is set for the block figure of 6. That is, in the width of the class having an area value of 500 to 600, the frequency is 2 (circle 32 and triangle 34), and the area value is 700 to 80.
By setting the frequency to 1 (square 36) for the width of the class of 0, the area histogram pattern of this inspection object is obtained.

Area histogram pattern creating means 2
In 0, the master pattern of the area histogram patterns of all the inspection objects that flow into the transport system is created in advance, and the identification means 22 uses the area histogram patterns for matching to inspect the inspection objects. I do.

Specifically, first, when creating the master pattern, as shown on the left side of FIG. 4, the image of each inspection object is input (step 110) and the binary image of the grayscale image is subjected to the normal binarization processing. (Step 120), and a block figure characteristic of the design of the object is extracted. Next, the area of each block figure obtained by the extraction processing is measured (step 130), and the master pattern of the area histogram pattern is created by the above method (step 140).

In the inspection process, it is determined whether or not a certain planned conveyed product has been conveyed normally. Specifically, as shown on the right side of FIG.
0), based on the master pattern, the width of the class of the histogram and the number of classes are set (step 220). Next, an inspection target image is input (step 230), binarization processing (step 240), and area measurement (step 250).
After that, an area histogram pattern of the input image is created (step 260), and the identification means 22 performs matching with the master pattern of the planned conveyed product (step 27).
0), a match / mismatch is determined (step 280).

As described above, the inspection is carried out by using the area histogram pattern. In this method, the accuracy of the master pattern is very important. , It may be indistinguishable from other inspection objects. Therefore, FIG. 5 shows a flow of processing when the master pattern is optimized.

In this processing, first, an image of a certain inspection object is input (step 310) and binarization processing is performed (step 320) to extract a block figure (step 330). At this time, if extraction of the block figure characteristic of the inspection object is insufficient, the binarization process is performed again (step 320). Subsequently, the area is measured (step 340), an area histogram pattern is created (step 350), and it is verified by the inspection processing shown on the right side of FIG. 4 whether the planned conveyed product can be completely distinguished from other inspection objects. Is performed (steps 360 and 370). As a result, when the distinction is not possible, the width of the class and the number of classes of the histogram are changed (step 380), and the area is again determined based on the block shape existing in the width of the class and the number of classes. A histogram pattern is created (step 350).

In this manner, the repeated verification is performed, and the completely distinguished area histogram pattern is adopted as the master pattern of the inspection object (step 39).
0).

If there is a very similar inspection object and it is indistinguishable from other inspection objects, the area value of the block figure existing only in another inspection object that cannot be indistinguished is It is possible to identify by creating an empty class that does not exist as a class of the histogram and creating the master pattern exclusively.

In the above embodiment, the area is used as the characteristic amount, but the type of the characteristic amount is not limited to this, and other characteristic amounts such as the perimeter and the circularity can be used. is there. It is also possible to perform the inspection by combining the area and the other feature amount.

[0029]

As described above, according to the present invention, a feature amount histogram pattern is created according to the characteristics of the pattern of the inspection target object, so that the inspection target object is prevented from being rotated or displaced due to the influence of the transport system. Even if it occurs, it has an excellent effect that a high recognition accuracy and high speed can be guaranteed by a simple method.

[Brief description of drawings]

FIG. 1 is a block diagram showing a basic configuration of an embodiment of an image processing inspection device according to the present invention.

FIG. 2 is a diagram showing an example of an inspection target object.

FIG. 3 is a diagram showing an example of an area histogram pattern created corresponding to the inspection object of FIG.

FIG. 4 is a flowchart showing a flow of creation of a master pattern using an area histogram pattern and inspection processing in the embodiment.

FIG. 5 is a flowchart showing an optimization process of a master pattern of an area histogram pattern in the above embodiment.

[Explanation of symbols]

 Reference numeral 10 ... Imaging camera 12 ... Capture circuit 14 ... Image memory 16 ... Extraction circuit 18 ... Area measurement circuit 20 ... Area histogram pattern creating means 22 ... Identification means 30 ... Inspection object 32, 34, 36 ... Bulk figure

Claims (4)

[Claims] 1. An image processing inspection device for inspecting a known object having a characteristic pattern during transportation, an image pickup means for imaging the inspected object, and an image memory in which the image data taken is captured. And an extracting means for extracting a characteristic block figure included in the inspection object from the image data on the image memory, and a feature amount measuring means for measuring a feature amount of each extracted block figure. A feature amount histogram pattern creating unit that creates a feature amount histogram pattern of the inspection target object by combining the feature amount values, and an identification unit that identifies the inspection target object by using the created feature amount histogram pattern are provided. An image processing inspection device characterized by the above. 2. The image processing inspection apparatus according to claim 1, wherein the class width and the number of classes of the feature amount histogram pattern are changeable. 3. The image processing inspection according to claim 2, wherein the class width and the number of classes of the feature amount histogram pattern are corrected according to the discrimination result of the inspection object and other objects. apparatus. 4. The image processing inspection device according to claim 1, wherein two or more types of feature quantities of each block figure are used in combination.