JPH08189902A - Picture processing device - Google Patents

Abstract

PURPOSE: To easily and quantitatively measure the texture of variable density picture at an interval nearly the same as visual inspection with small calculation quantity based on the edge information of an object to be tested with gray level picture. CONSTITUTION: A steel plate 1 as a transfer object to be tested is illuminated by an illumination 2 and deffectives 3 on the surface thereof such as scratch, recessed part, etc., are picked up by a camera 4, and the surface picture of the plate 1 is A/D converted by a signal processing circuit 5, then the output picture data is stored by a picture memory 6. The optional rectangular area in the picture data from the memory 6 is set by a mask setting circuit 7 and it is differentiated by a differentiating circuit 8, then a characteristic quantity calculation circuit 9 calculates a sum or squared sum of differentiation information with a specified value or more regarded as edge of defect in the rectangular area and the number of picture elements of the differentiation information. The sum or squared sum thereof and the number thereof show the intensity and contrast of the edge in the rectangular area and the number of edges respectively, and a circuit 10 for judging the presence or absence of defect and the degree of existence judges the presence or absence of defect and the degree of existence thereof on the surface of the plate 1 based on the obtained results.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a technique for quantifying the texture of a grayscale image based on the edge information of the grayscale image object, for example, surface defects of a product such as a steel plate carried out from a production line. The present invention relates to an image processing apparatus for performing image processing on a grayscale image of a surface of the image in order to inspect.

[0002]

2. Description of the Related Art In surface defect inspections of various production lines, it is necessary to evaluate not only the presence or absence of defects but also how clearly and how many defects are visible. However, the visual evaluation by human eyes varies from person to person, and there is a subjective variation, which puts aside on objective quantitativeness. Therefore, conventionally, various quantification methods of grayscale images and calculation methods of their feature amounts have been proposed.

For example, the document “Adaptation of Texture Analysis to Scintigrams” (Radioisotopes V
ol, 40, No1 pp. 33-36 1991), image processing is performed on a scintigram of a lesion site read by a doctor, and a visual evaluation of the scintigram is performed based on five types of feature values obtained from the image: average, variance, skew, energy, and entropy. We do quantitatively. In particular, the variance is conventionally often used to represent the variation state of the brightness value, but when the shading or the like exists in the image, the variance feature amount is not desirable. In addition, analysis is performed by using a density co-occurrence matrix, density difference statistics, density run length matrix, etc. in the method commonly used for texture analysis (Image Analysis Handbook (The University of Tokyo Press) pp. 518-522). There are methods, and among these methods, it is known that the method using the concentration co-occurrence matrix is effective. The density co-occurrence matrix, the density difference statistic are second-order statistics of pixel density, and the density run-length matrix is a higher-order statistic of pixel density. A feature amount that can be calculated from these statistics is defined, and analysis such as classification and identification is performed using this feature amount (JP-A-5-146433, JP-A-6-63).
-60182).

[0004]

However, these generally have a large amount of calculation, and the calculation result differs depending on the vector set at the time of calculation, and it is difficult to determine appropriate parameters. In addition, if the matrix is made large to increase the amount of information, the amount of calculation becomes very large. Therefore, it is necessary to determine an appropriate matrix size.

In view of the above circumstances, it is an object of the present invention to provide an image processing apparatus that can easily quantify the texture of an image with a calculation amount smaller than that of the texture analysis described above and with a feeling close to that of visual observation. To do.

[0006]

An image processing apparatus that achieves the above object includes (1) an image data storage unit for storing image data representing a grayscale image, and (2) a grayscale image stored in the image data storage unit. Mask setting unit that cuts out a partial area (3) Differential operation unit that generates differential image data by differentiating the image data representing the partial area cut out by the mask setting unit (4) Differentiation generated by the differential operation unit An index calculation unit is provided for determining a first index indicating the degree of variation of image data and a second index indicating the number of pixels having differential image data exceeding a predetermined value in the differential image data. It is what

Here, the above-mentioned differential operation section performs a differential operation in each of the two directions intersecting with each other and generates differential image data for each pixel based on the two data obtained for each pixel. It is effective to have.
Further, the index calculation unit obtains, as the first index, the sum of the differences between the differential image data and the predetermined value or the sum of squares of the differential image data that exceeds a predetermined value among the differential image data. It is preferable.

[0008]

The image processing apparatus of the present invention obtains the above-mentioned first index and the above-mentioned second index based on the differential image data, and the defect on the grayscale image is detected by the first index. The degree of how clear the image looks can be known, and the second index shows how many defects are present on the grayscale image. As will be described later, the first index and the second index enable objective evaluation close to human visual inspection.

[0009]

Embodiments of the present invention will be described below. FIG. 1 is a block diagram of an embodiment of the image processing apparatus of the present invention. A steel plate 1 as a transported inspection object is illuminated by a light 2. The surface of the steel plate 1 has defects 3 such as scratches and dents. The surface image of the steel plate 1 taken by the camera 4 is A / D converted by the signal processing circuit 5, and the output image data is stored in the image memory 6. The mask setting circuit 7 sets an arbitrary rectangular area from the image data read from the image memory 6, and the differentiation processing circuit 8 performs the differentiation processing.

FIG. 2 is an explanatory diagram of the differential processing in the differential processing circuit shown in FIG. As shown in FIG. 2, this differentiating process is performed by centering this pixel on each pixel of the image to be processed.
The differential filter F is superimposed on the × 3 region, the product of the corresponding pixels is obtained, and the sum of them is used as the differential value. This operation is performed from the upper left image to the lower right pixel.

Although it is not necessary to change the setting of the differential filter F acting on the grayscale image according to the shape of the object, a differential filter for extracting edges in only a certain direction can find only the edge of the object in that direction. . It suffices if the target object can be detected, but in general, the accuracy in detecting the target will be reduced. Therefore, the differential filter F is preferably a Sober filter that detects edges in all directions and is not easily affected by background noise, or a differential filter that operates similarly to it. However, if a differential filter that detects edges with a certain directionality is applied, the texture can be quantified for that directionality. Therefore, if the type or directionality of the defect is known, there is a certain directionality. A differential filter for detecting edges may be operated.

FIG. 3 is a diagram showing a Sober filter as the differential filter shown in FIG. The Sober filter shown in FIG. 3 is composed of an X-direction filter (a) having a weighting factor in the X direction and a Y-direction filter (b) having a weighting factor in the Y direction. If such a Sober filter is used as the differential filter used for the differential processing,
It is possible to perform not only the edge in one direction but also the edge in all directions to perform the differential processing in which the influence of the background noise is suppressed. In this case, the result of applying the X-direction filter (a) is Δ
Δxy is the differential value, where Δy is the result of the action of the x, y direction filter (b). Here, the differential value Δxy
Is expressed as ^{Δxy = √ (△ x 2 +} △ y 2).

The differential value Δxy thus obtained represents the rate of change of the brightness value in the area near the pixel. There are Laplacian, Roberts, Prewitt, Kirsch filters, etc. that are often used in addition to the Sovert filter. However, if the sum of the weighting factors in the filter is 0, the size, weight, and coefficient may be set arbitrarily. . Also,
You may combine several filters.

Returning to FIG. 1 again, the description will be continued. The feature amount calculation circuit 9 calculates the sum or squared sum of differential information of a predetermined value or more that can be regarded as an edge of a defect in a rectangular area and the number of pixels of differential information of a predetermined value or more that can be regarded as an edge of a defect in a rectangular area. To do. The sum or square sum of the differential information above a predetermined value that can be regarded as a defect edge indicates the strength and contrast of the edge in the rectangular area, and the number of pixels of the differential information above the predetermined value that can be regarded as a defect edge is the edge. The presence / absence determination circuit 10 determines the presence / absence and presence of a defect on the surface of the steel sheet 1 based on these.

4 and 5 show the presence or absence of defects on the surface of the steel plate 1,
It is a determination diagram of the degree of existence. Defect presence / absence determination circuit 10
Is the presence or absence of defects on the surface of the steel plate 1 in the range as shown in FIG.
Existence is judged. Generally, steel sheets are divided into five stages from those in which no defects are visible to those in which a large number of defects are clearly visible.

As shown in FIG. 5, the degree of defects, that is, the degree of existence is quantitatively classified into five levels from the calculated feature amount and a predetermined threshold value. FIG. 6 is a diagram showing comparison data of numerical data of defects on the surface of the steel sheet 1 and visual observation results. The numbers in parentheses () next to the plots shown in FIG. 6 are the numbers given in the order of the defects when visually inspected.
Although there were some differences among individuals, the results were close to the visual sense as shown in FIG.

[0017]

As described above, the image processing apparatus of the present invention is particularly effective for calculating the degree of presence close to the visual sense, which indicates the severity of defects such as planar defects, and the image shading. The texture of an arbitrary area can be easily quantified with a small amount of calculation as compared with the conventional method without considering the above.

[Brief description of drawings]

FIG. 1 is a block diagram of an embodiment of an image processing apparatus of the present invention.

FIG. 2 is an explanatory diagram of a differential process in the differential processing circuit shown in FIG.

FIG. 3 is a diagram showing a Sober filter as the differential filter shown in FIG. 2;

FIG. 4 is a diagram showing the presence / absence of defects on the surface of a steel sheet and the degree of presence thereof.

FIG. 5 is a diagram showing an example of determining the presence / absence of defects on the surface of a steel sheet.

FIG. 6 is a diagram showing comparison data of numerical data of defects on the surface of a steel sheet and visual observation results.

[Explanation of Codes] 1 Steel plate 2 Illumination 3 Defect 4 Camera 5 Signal processing circuit 6 Image memory 7 Mask setting circuit 8 Differentiation processing circuit 9 Feature amount calculation circuit 10 Defect presence / absence determination circuit

Claims (3)

[Claims] 1. An image data storage unit that stores image data representing a grayscale image, a mask setting unit that cuts out a partial region of the grayscale image stored in the image data storage unit, and a portion cut out by the mask setting unit. A differential operation unit that generates differential image data by differentiating image data that represents a region, a first index that indicates the degree of variation in the differential image data that is generated by the differential operation unit, and among the differential image data An image processing apparatus comprising: an index calculation unit that obtains a second index that represents the number of pixels having differential image data that exceeds a predetermined value. 2. The differential operation section performs a differential operation in each of two intersecting directions, and generates differential image data for each pixel based on the two data for each pixel obtained by the differential operation. The image processing apparatus according to claim 1, wherein 3. The index calculation unit obtains, as the first index, a sum or a square sum of differences between the differential image data and the predetermined value for the differential image data that exceeds a predetermined value among the differential image data. The image processing apparatus according to claim 1, wherein the image processing apparatus is an image processing apparatus.