JPH0886759A - How to identify surface defects - Google Patents

Abstract

PURPOSE: To attain the discrimination of the kind of an irregularity defect with high accuracy by calculating a feature amount showing the feature of the defect, and discriminating the defect according to the feature amount. CONSTITUTION: An image signal is obtained by photographing the illuminated surface of a subject to be inspected. In accordance with the image signal, a defective region corresponding to a defect in the surface of the subject to be inspected is extracted from within an image showing the surface of the subject to be inspected, and average values of image signals corresponding to the overlapping portions of the defective region with a plurality of partial regions obtained by the dividing of the inside of the rectangular region into plural (three) parts by one side of the contour of the rectangular region, e.g. a line parallel to the Y axis, are calculated. The plurality of average values are digitized by comparison of the plurality of average values with one or plural thresholds. A feature amount showing the feature of the defect is calculated on the basis of a pattern of arrangement of the plurality of these numerical values, and the defect is identified according to the feature amount.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention provides an image pickup signal by picking up an image of the surface of a strip-shaped member to be inspected, which moves at high speed, such as a steel plate, by an image pickup device such as a CCD camera. The present invention relates to a surface defect discriminating method for discriminating the surface defect.

[0002]

2. Description of the Related Art Conventionally, as one of image processing apparatuses for defect type determination which employs the above-described surface defect determination method, for example, an image processing apparatus for surface defect determination manufactured by ISYS, Inc. in the United States is known. In this image processing device, dozens of types of feature amount calculations are performed on the defect portion extracted by the defect extraction circuit to obtain the feature amount, and the value of the feature amount of the defect discrimination table registered in advance is calculated. And the feature amount obtained by the calculation of the extracted defects are compared,
When the feature amount of the extracted defect portion matches the condition of the value range of the feature amount registered in the defect determination table, the defect is the defect type registered in the table. It is a judgment.

To exemplify the feature amount referred to at this time, for example, the shape of the extracted defect is close to a circle, elongated in the vertical direction (y direction), or elongated in the horizontal direction (x direction). Represents the feature amount, which represents the defect size such as the length, width, perimeter, and area of the extracted defect, and the brightness such as the maximum value, the minimum value, and the average value of the grayscale image of the extracted defect portion. There are characteristic quantities.

Further, CAT-BO, which is an image processing apparatus used for determining the type of surface defect of a steel line of Toshiba Corporation.
X applies the detected defect feature amount to a tree-structure determination logic that is built in advance, and determines the defect type according to the logic. This Toshiba CAT-BO
The feature amount referred to by X is also of the same type as ISYS described above.

[0005]

As described above, various characteristic amounts have been conventionally obtained and defects are discriminated based on these characteristic amounts. For example, a defect of unevenness on the surface of a steel sheet is There is a problem that extremely close feature amounts are often required regardless of the difference in the cause of the occurrence, and the type of the uneven defect is often erroneously determined.

In view of the above circumstances, the present invention is
It is an object of the present invention to provide a surface defect discriminating method capable of discriminating types of irregularities with high accuracy.

[0007]

According to a first discrimination method of the present invention for achieving the above object, an image of an illuminated surface of an object to be inspected is obtained to obtain an image pickup signal, and the surface of the object to be inspected is imaged based on the image pickup signal. In the surface defect determining method for determining a defect, a defect area corresponding to a defect on the surface of the object to be inspected is extracted from an image representing the surface of the object to be inspected based on the image pickup signal, and the inside of the defect area is divided into a plurality of areas. The average value of the image pickup signal corresponding to each of the plurality of partial areas is obtained, the feature amount representing the feature of the defect is obtained based on these average values, and the defect is identified based on the feature amount. It is characterized by making a distinction.

The first discriminating method of the present invention may be configured concretely as long as the above requirements are satisfied. For example, as one embodiment, it may be constructed as follows. That is, a surface defect discriminating method as one aspect of the first discriminating method of the present invention captures an image of an illuminated surface of an inspected object to obtain an imaging signal, and detects a defect on the surface of the inspected object based on the imaging signal. In the method of determining the surface defect to be determined, based on the image pickup signal, in the image representing the surface of the inspection object,
A defect area corresponding to a defect on the surface of the object to be inspected is extracted, a rectangular area circumscribing the defect area is obtained, and the inside of the rectangular area is divided into a plurality of lines by a line segment parallel to one side of the outline of the rectangular area. Obtaining each average value of the image pickup signal corresponding to each overlapping portion of each of the plurality of partial areas and the defect area, and comparing each of these plurality of average values with one or more threshold values In this way, each of the plurality of average values is digitized, the feature amount representing the feature of the defect is obtained based on the array pattern of the plurality of numbers, and the defect is discriminated based on the feature amount. To do.

Further, the second judging method of the present invention for achieving the above object obtains an image pickup signal by picking up an image of the illuminated surface of the object to be inspected, and judges a defect on the surface of the object to be inspected based on the image pickup signal. In the method of discriminating a surface defect, a defect area corresponding to a defect on the surface of the object to be inspected in an image representing the surface of the object to be inspected is extracted based on the image pickup signal, and on a predetermined straight line passing through the defect area. The feature is that a profile of an image pickup signal is obtained, a feature amount representing a feature of a defect is obtained based on this profile, and a defect is discriminated based on this feature amount.

Regarding the above-mentioned second discrimination method of the present invention,
Similar to the first determination method, any specific configuration may be used as long as the above requirements are satisfied. For example, the configuration may be as follows in one embodiment. That is, a surface defect discriminating method as one embodiment of the second defect discriminating method of the present invention is to capture an imaged signal of an illuminated surface of an inspected object and obtain an image pickup signal based on the imaged signal. In the method of determining a surface defect, a defect area corresponding to a defect on the surface of the object to be inspected in an image representing the surface of the object to be inspected is extracted based on the image pickup signal, and a representative inside the defect area is extracted. The point is obtained, the profile of the image pickup signal on a straight line passing through the representative point is obtained, the number of times the profile intersects a predetermined threshold level, and the difference between the profile and the threshold level are calculated by the straight line. It is characterized in that the feature amount representing the feature of the defect is obtained based on the sign of the initial difference when sequentially obtained along with, and the defect is discriminated based on the feature amount.

[0011]

The gray scale image of the surface defect portion of the steel sheet, which is picked up by the image processing apparatus by picking up an image from the regular reflection direction or the irregular reflection direction using a CCD camera as the detection head, will be described. FIG. 1 is a conceptual diagram of a light reflection pattern of an uneven defect such as an oyster ball portion of an oyster defect or an indentation defect defect portion in which an edge is steep to some extent. The solid line represents the reflected light intensity distribution in the presence of the uneven defect, and the broken line represents the reflected light intensity distribution of a normal surface in which no defect exists. The same applies to FIG. 4 referred to below.

In these concave and convex defects, the grayscale image of the defective portion taken from the direction of specular reflection is darker as a whole, as shown in FIG. However, as shown in FIG. 3, the grayscale image of the defect portion imaged from the irregular reflection direction in FIG. 1 shows that the edge portion of the defect on the side opposite to the incident light side is brighter than the defect-free portion, Other parts tend to be dark.

On the other hand, the edge changes gently,
FIG. 4 is a conceptual diagram of a light reflection pattern of an uneven defect having a rougher surface than the normal part over the entire defective part, for example, a hedging defect. As shown in FIG. 5, the grayscale image obtained by capturing the irregularity defect from the regular reflection direction is darker in the defect portion as a whole as compared with the normal portion having no defect, and is similar to the grayscale image in the case of FIG. Shows the pattern. On the other hand, as shown in FIG. 6, the grayscale image of the defective portion imaged from the direction of irregular reflection in FIG. 4 is a brighter image as a whole than the portion without the defect.

Focusing on FIGS. 3 and 6, it can be seen that there is a difference in the light and dark patterns of the grayscale image picked up from the irregular reflection direction between the unevenness defect having a sharp edge and the unevenness defect having a gentle edge. The present invention has been completed paying attention to this point. That is, in the first determination method of the present invention, each average value of the image pickup signal corresponding to each of the plurality of partial areas obtained by dividing the inside of the defective area into a plurality of areas is obtained, and these average values are obtained. Since the feature amount representing the feature of the defect is obtained based on the feature amount and the defect is determined based on the feature amount, the defect type is determined by the cause of the uneven defect, that is, the uneven defect having a sharp edge and the edge It is possible to discriminate a smooth uneven defect very easily and without error.

Further, according to the second discrimination method of the present invention, the profile of the image pickup signal on the straight line passing through the defect area is obtained, the feature amount representing the feature of the defect is obtained based on this profile, and based on this feature amount. Since the defect is discriminated by the above-described method, similar to the first defect discrimination method, the uneven defect is discriminated with high accuracy according to the cause.

[0016]

Embodiments of the present invention will be described below. FIG. 7 is a schematic view of a surface defect discrimination system adopting an embodiment of the method for discriminating surface defects according to the present invention, and shows a state in which the surface of a steel sheet moving at high speed is inspected.

Power is supplied to the light source 1 from the power source 10 for the light source, the light source 1 irradiates the inspected material 2 with the light beam 3, and the irradiation portion 4 of the inspected material 2 is imaged by the one-dimensional CCD camera 5 from the irregular reflection direction. . The one-dimensional signal output from the one-dimensional CCD camera 5 is A / D-converted at high speed by the high-speed A / D converter 6, and the signal is input to the defect extraction signal processor 7 to extract the defect signal.

In the defect extracting signal processing device 7, while the shading correction (correction of the signal strength difference between the end portion and the central portion during image pickup) is performed on the input digital signal, the corrected digital signal is Sequentially stored in the image memory provided inside the defect extraction signal processing device 7,
As a result, the two-dimensional image is stored in this image memory.
A grayscale image of the defective portion is extracted by extracting a signal exceeding a predetermined threshold value as a defective portion signal from the digital image pickup signal representing the two-dimensional image.

A method of calculating the characteristic quantities A and B will be described below as an embodiment of the present invention. <Calculation Method of Feature A> (1) The defect extraction signal processing device 7 extracts a grayscale image of the defect portion.

The signal of this defective portion is input to the image processing apparatus 8 for defect determination, and the image processing apparatus 8 for defect determination executes the following steps (2) to (4). (2) A rectangle circumscribing the grayscale image of the defective portion is generated. (3) Divide the above rectangle parallel to one side of the rectangle. (4) The average gray scale level v of the defective portion in each of the plurality of subdivided rectangles is calculated based on the following equation (1).

[0021]

[Equation 1]

Here, N represents the total number of pixels in the defective portion in each subdivided rectangle, and p _i represents the gray scale level value of the i-th pixel in the defective portion in each subdivided rectangle. (5) The defects in the divided rectangles are digitized from the value of the average gray scale level v. (6) The numerical value array pattern is converted into a numerical value to obtain the feature value A.

The steps (2) to (6) will be specifically described below. FIG. 8 is a schematic diagram of a grayscale image of an extracted defect portion of a concave-convex defect having a sharp edge. Here, a rectangle circumscribing the grayscale image of the extracted defect portion is generated, and as an example, y
It is divided into three equal parts parallel to the axis. The average gray scale level v of the defective portion in the subdivided rectangle is set to the above (1).
Calculate according to the formula. The grayscale level is set to, for example, 256 levels. For example, when binarizing the grayscale level, the grayscale level v = 128 is set as a reference grayscale level without defects, and the grayscale level v is 128 or more (the reference scale level is It is set to 1 when it is brighter and 0 when the gray scale level is less than 128 (when it is darker than the reference scale level).

When the rectangle is divided into three and the gray scale levels are binarized, there are 2 ³ = 8 gray scale level arrangement patterns. The grayscale level array pattern of the defects in FIG. 8 is represented by (101). If this is expressed by the decimal-based feature quantity, it becomes 7. FIG. 9 is a schematic diagram in which a grayscale image of the same defect is divided into three equal parts in parallel with the x-axis, and the grayscale level array pattern of the defect portion is obtained in the same manner as above (100). This is 6 when expressed in decimal notation.

In the case of a defect having a gentle edge, the array pattern of the gray scale level of the defect portion is (111) in both cases when divided in the y-axis direction or the x-axis direction.
If this is expressed as a feature amount, it becomes 8. It is possible to discriminate because a defect having a sharp edge and a defect having a smooth edge have different feature amounts.

<Calculation Method of Feature B> As shown in FIG. 10, even in the case of a defect having a sharp edge,
Since the area brighter than the surrounding area is small, and the feature quantity that there is a defect brighter than the surrounding area cannot be obtained by the above-described method of calculating the feature quantity A, the defect is discriminated from the grayscale profile. Add method. First, the coordinates of the center of gravity of the extracted defective portion are calculated.

FIG. 13 is a schematic diagram for explaining how to find the center of gravity. Each square corresponds to each pixel. The coordinates (x ₀ , y ₀ ) of the center of gravity are obtained from the calculation formula shown in Formula (2) with respect to the coordinates of the extracted defect portion in the x direction and the y direction as shown in FIG.

[0028]

[Equation 2]

However, N is the total number of pixels of the extracted defective portion, and x _i and y _i are the x and y coordinates of the i-th pixel of the extracted defective portion. After obtaining the coordinates of the center of gravity, a profile of gray scale values in the y direction passing through the center of gravity (x ₀ , y ₀ ) is obtained as shown in FIG. 11, and the number of times the profile intersects the reference gray scale level is obtained. . The determination of the intersection depends on whether or not the grayscale level of the adjacent pixel crosses the reference grayscale level.
Further, the starting point of the profile is determined in advance (in this embodiment, the incident light side is the reflecting side), and when the gray scale level of the starting point is higher than the reference gray scale level,
A positive + sign and a negative − sign when the value is low are added to the obtained number of crossings, and this value is set as the feature amount B. In the case of FIG. 11, +1 is the value of this feature amount B. In the case of FIG. 12 having a complicated shape, since the profile is in contact with the reference gray scale level but does not cross it, the points a and b are not considered to intersect, and the feature amount B is +2. .

Although the method for determining the feature amount of the gray scale value profile in the y direction has been described above, the feature amount of the profile of the gray scale value in the x direction may be determined for determination. In the case of a smooth edge defect, the number of intersections is 0 and the sign is +, so the feature amount B is +0. Since the features of a defect having a sharp edge and a defect having a gentle edge are different, it is possible to easily discriminate them.

The characteristic amounts A and B obtained as described above are
By using it together with the feature quantity representing the shape, the feature quantity representing the size, the density feature quantity, etc., which have been conventionally used, it has been difficult to discriminate using only the conventionally used feature quantity. It is possible to very easily discriminate between two types of concave and convex defects, that is, a convex and concave defect and a concave and convex defect with a smooth edge.

Table 1 is an example of a feature amount registration table. In this table, "heavy" as an example of a defect having a gradual edge and an example of a defect having an abrupt edge are shown. The feature type for defect type determination for “Oyster defect” is registered. It is possible to register the maximum value (max) and the minimum value (min) for each feature amount, and the extracted defect feature amount falls within the range of the defect feature amount values registered in this table. If they match, the defect will be determined. Registration of feature quantity is sample test,
Alternatively, a large number of flaw data of various flaw types are collected by an online test, and the range of the value of the feature amount is determined from the flaw data.

Here, a cold-rolled steel sheet having about 20 coils was used to inspect each coil for a length of 300 m to 400 m at a line speed of 300 m / min by a system shown in FIG. Regarding "ga", the defect was identified by the registration table of the feature amount of Table 1. On the other hand, these coils were rewound at a low speed, and the operator visually observed the defect at the same position as the defect detected by the system shown in FIG. 7 to determine the defect. For each defect,
A value obtained by dividing the number of defects determined to be oyster flaws by the system of FIG. 7 and the defects of which are also visually determined by an operator by the number of defects determined to be oyster flaws in the system of FIG. The rate of agreement was "Oyster flaw". The concordance rate of "heard" was calculated in the same manner.

Table 2 shows the coincidence rate of defects when a defect is discriminated only by the conventional feature amount, and when the discrimination by the conventional feature amount is added to the discrimination by the feature amount according to the present embodiment. When the determination according to the present embodiment is added to the determination based on the conventional feature amount, the determination means that a logical product of a result determined based on the conventional feature amount and a result determined based on the feature amount according to the present embodiment (each Also if it is determined to be defective, it is determined to be defective). In Table 2, and represents a logical product, and or represents a logical sum.

When the defect is discriminated by adding the feature amount of this embodiment, it is understood that the defect coincidence rate is significantly improved as compared with the conventional defect discrimination using only the feature amount.

[0036]

[Table 1]

[0037]

[Table 2]

[0038]

As described above, by adopting the method for discriminating surface defects of the present invention, it is possible to discriminate uneven defects with high accuracy.

[Brief description of drawings]

FIG. 1 is a conceptual diagram of a light reflection intensity distribution with respect to a concave-convex defect having a sharp edge.

FIG. 2 is an example of a gray scale image when a concave-convex defect having a sharp edge is imaged with a detection head placed in the regular reflection direction.

FIG. 3 is an example of a gray scale image when a concave-convex defect having a sharp edge is imaged with a detection head placed in the irregular reflection direction.

FIG. 4 is a conceptual diagram of a light reflection intensity distribution with respect to a concave-convex defect having a smooth edge.

FIG. 5 is an example of a grayscale image when a concave-convex defect having a smooth edge is imaged with a detection head placed in the regular reflection direction.

FIG. 6 is an example of a grayscale image when a concave-convex defect having a smooth edge is imaged with a detection head placed in the irregular reflection direction.

FIG. 7 is a schematic view of a surface defect discrimination system adopting an embodiment of a surface defect discrimination method according to the present invention.

FIG. 8 is a diagram showing an example in which a defect image is divided in parallel in the y direction in order to calculate a characteristic amount A.

FIG. 9 is a diagram showing an example in which a defect image is divided in parallel in the x direction in order to calculate a feature amount A.

FIG. 10 is a diagram showing an example of a case in which a feature amount A of the present invention cannot identify a defect.

11 is a schematic diagram for calculating a feature amount B. FIG.

FIG. 12 is a diagram showing how to determine the center of gravity of a defective portion.

FIG. 13 is a schematic diagram when the defect shape for calculating the feature amount B is complicated.

[Explanation of symbols]

 1 Light Source 2 Inspected Material 5 One-dimensional CCD Camera 6 A / D Converter 7 Defect Extraction Signal Processing Device 8 Defect Extraction Signal Processing Device 9 Display Device

Claims (4)

[Claims] 1. A method for determining a surface defect in which an illuminated inspection object surface is imaged to obtain an imaging signal, and a defect on the inspection object surface is determined based on the imaging signal. , A defect area corresponding to a defect on the surface of the object to be inspected in the image representing the surface of the object to be inspected is extracted, and an image pickup signal corresponding to each of a plurality of partial areas in which the inside of the defect area is divided A method for determining a surface defect, characterized in that each average value is obtained, a feature amount representing a feature of a defect is obtained based on the plurality of average values, and the defect is identified based on the feature amount. 2. A surface defect determining method for determining a defect on the surface of an object to be inspected on the basis of the image signal by imaging an illuminated surface of the object to be inspected, and based on the image signal. , A defect area corresponding to a defect on the surface of the object to be inspected in the image representing the surface of the object to be inspected, a rectangular area circumscribing the defect area is obtained, and a line parallel to one side of the contour of the rectangular area Each of the plurality of partial areas obtained by dividing the inside of the rectangular area into a plurality of areas by the minute, and the average value of the imaging signal corresponding to each of the overlapping portions of the defect area are obtained, and the average values of the plurality of average values are calculated. Each of these average values is digitized by comparing each with one or a plurality of threshold values, and the feature quantity representing the feature of the defect is obtained based on the array pattern of these plurality of values. , Method of determining surface defects, characterized in that to determine the defect based on the feature quantity. 3. A surface defect determining method for determining a defect on the surface of an object to be inspected based on the imaged signal by imaging an illuminated surface of the object to be inspected, and based on the image capturing signal. , A defect area corresponding to a defect on the surface of the object to be inspected in an image representing the surface of the object to be inspected is extracted, and a profile of an imaging signal on a predetermined straight line passing through the defect area is obtained, and based on this profile A method of determining a surface defect, characterized in that a feature quantity representing a feature of a defect is obtained, and the defect is identified based on the feature quantity. 4. A surface defect determining method for determining a defect on the surface of an object to be inspected based on the imaged signal by imaging an illuminated surface of the object to be inspected, the method comprising: , A defect area corresponding to a defect on the surface of the inspected object in the image representing the surface of the inspected object is extracted, a representative point inside the defective area is obtained, and an imaging signal on a straight line passing through the representative point is extracted. Defects based on the number of times the profile intersects a predetermined threshold level and the sign of the initial difference when the difference between the profile and the threshold level is sequentially determined along the straight line. A method of determining a surface defect, which comprises: determining a feature amount representing the feature of 1. and determining a defect based on the feature amount.