JP3255562B2 - Surface inspection equipment - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a surface inspection apparatus for detecting a surface defect of a running strip.

[0002]

2. Description of the Related Art FIG. 11 shows an example of a conventional surface inspection apparatus.

As shown in FIG. 11, the surface inspection apparatus comprises:
A detector (detection means) 3 for detecting an image of a surface defect 2 of the running strip 1 and a roller 4 for detecting the running of the strip 1
A line synchronizing signal generator 5, a dividing circuit 6 for dividing the inspection target surface of the strip 1 into divided surfaces of a fixed size in the width direction and the longitudinal direction irrespective of a defect occurrence position, An identification circuit 7 for determining the type and grade of a defect included therein is provided, and the surface of the strip 1 is automatically inspected.

FIG. 12 shows an example of the occurrence of a defect on the surface of the strip 1 and an example of the configuration of the division surface. In the figure, A is a linear defect distributed in the longitudinal direction of the strip 1, B is an aggregate of point defects, C is an intermittent thin linear defect, and D is an isolated point defect. L ₁ , L ₂ , L ₃ ,... Are longitudinal dividing lines, and D ₁ , D ₂ , D ₃ are dividing lines in the width direction. Divided into division planes.

[0005] In the conventional surface inspection apparatus, the inspection target surface of the strip is divided into divided surfaces of a fixed size, and the type and grade of the defect are determined for each defect included in each divided surface.

[0006]

However, as shown in FIG. 12, many surface defects of a belt-like material are to be determined as one defect by collecting several defects, and the shape of the defect is generally one. It is complicated and the position of occurrence is undefined.

For this reason, since the inspection target surface of the strip is divided into divided surfaces of a fixed size in the width direction and the longitudinal direction, for example, the defects A, B, and C straddle a plurality of divided surfaces. There is a problem that it is difficult to determine the type and grade correctly.

As shown in FIG. 6A, when one defect is composed of a hatched dark portion and a dotted light portion, the binarization is performed as shown in FIG. As shown in FIG. 7, since a dark portion remains, it may be separated into a plurality of defective portions.

Further, as shown in FIGS. 7A and 7B,
If the entire defect is faint (light), binarization almost eliminates the defective portion and makes it impossible to determine the defect area. Therefore, it is difficult to correctly determine the type and grade of the defect.

[0010] Therefore, an object of the present invention is to make it difficult to determine a defective portion when there are surface defects of various shapes on the strip and the surface defect is composed of a plurality of portions, or when the defect is faint (light). In any case, it is an object of the present invention to provide a surface inspection apparatus capable of quickly and accurately identifying the type and grade of each defect in real time.

[0011]

According to a first aspect of the present invention, there is provided a detecting means for detecting an image of a surface defect of a running belt, and the surface detected by the detecting means. Prior to binarizing a defect image, a local image memory for storing an image in a range to be subjected to multi-value image processing, a filter memory for storing parameters for multi-value image processing, a local image memory and a filter memory A multi-valued image filter circuit for performing multi-valued image processing with an arithmetic circuit for performing a multi-valued image calculation between the two, and a binarization circuit for binarizing the multi-valued image-processed image. Valued 2
A defect area determination unit including a second cutout circuit that determines a defect distribution state based on the pattern of the value image to determine a defect area composed of the same kind of defects; and a type of the defect included in the determined defect area. Identification means for identifying a class.

According to the first aspect of the present invention, the defect distribution state is determined from the image of the surface defect detected by the detecting means, so that a defect region surrounding the same type of defect such as a line or a dot can be formed. A determination is made and the type and grade of the defect in the defect area is identified. As a result, even when a surface defect having various shapes is present on the strip and the surface defect includes a plurality of portions, the type and grade of the defect can be quickly and accurately identified.

[0013]

According to the second aspect of the present invention, the defect area determination means includes a binarization circuit for binarizing the image of the surface defect,
For example, by using a crowd determination circuit that determines the same type of defect as one connection from the pattern of the binary image and a cutout circuit that determines each defect region by determining the separation of the defect determined as one connection, for example, A defect region surrounding the same type of defect, such as a line or a dot, is accurately determined.

[0015]

[0016]

[0017]

[0018]

[0019]

According to the first aspect of the present invention, the first extraction circuit takes the projections of the defects determined as one connection in the longitudinal direction and the width direction of the strip, and determines each defect area from these projections. It is also possible to use a projection system. When the projection method is adopted, it is possible to determine each defect area more easily and at a higher speed than the labeling processing.

[0021]

According to the first aspect of the present invention, in the above-mentioned projection system, the projection in the width direction is performed at each end of the projection of the defect in the longitudinal direction of the strip, and the rectangle determined by the corresponding projection in the longitudinal direction and the width direction. By determining each defect area in the area,
Simpler and faster processing becomes possible.

[0023]

[0024]

[0025]

[0026]

According to a second aspect of the present invention, there is provided a detecting means for detecting an image of a surface defect of a traveling belt-like object, and prior to binarizing the image of the surface defect detected by the detecting means.
A local image memory for storing images in a range to be subjected to multi-value image processing, a filter memory for storing parameters for multi-value image processing, and an operation for performing multi-value image operation between the local image memory and the filter memory A multi-valued image filter circuit for performing multi-valued image processing, a binarization circuit for binarizing the multi-valued image, and a defect from the pattern of the binarized binary image. Area determining means for determining a distribution state of the defects and determining a defect area composed of the same kind of defects, and identifying means for identifying the type and grade of the defect included in the determined defect area And characterized in that:

According to the second aspect of the present invention, in the multi-valued image filter circuit, an image in a range to be subjected to the multi-valued image processing is stored in the local image memory, and a filter function for performing the multi-valued image filtering is provided. Is stored in the filter memory in advance, and the output between the local image memory and the filter memory is input to the arithmetic circuit, and the "blur"
By performing the processing and the “integration” processing, it is possible to accurately determine a defective area when binarizing.

According to the second aspect of the present invention, the multi-valued image filter circuit of the defect area judging means receives an image of a surface defect and changes the image to a gradual change when the density of the defective portion is sharp. In addition, when the density of the defective portion is low, an "integration" process of floating the defective portion by integrating the images in an appropriate area is performed. The processed image is converted into a binary image by a binarizing circuit, and a defect region is determined from a pattern of the binary image by a second extracting circuit.

[0030]

[0031]

[0032]

[0033]

[0034]

[0035]

[0036]

According to a third aspect of the present invention, there is provided a detecting means for detecting an image of a surface defect of a traveling strip, and prior to binarizing the image of the surface defect detected by the detecting means,
A local image memory for storing an image in a range to be subjected to multi-valued image processing, a filter memory for storing parameters for multi-valued image processing, an output of the local image memory and an output of the filter memory; A multi-valued image filter circuit for performing multi-valued image processing including an arithmetic circuit for performing a multi-valued image operation by performing a product-sum calculation, a binarization circuit for binarizing the multi-valued image-processed image, A defect area determining means including a second cutout circuit for determining a defect distribution state from the pattern of the binarized binary image and determining a defect area composed of defects of the same type; and the determined defect area. And identification means for identifying the type and grade of the defect included in the information.

According to the third aspect of the present invention, in the arithmetic circuit, the image level of the address (I, J) of the local image memory is set to f (I, J), and the address (i, J) of the filter memory is set. The output of j) is g (i, j), the level of the address (I, J) of the image after the calculation is F (I, J), and the filter area is −l ₁ ≦ l ≦ l ₂ , −m ₁ ≦ m ≦ m ₂
When

(Equation 1) By performing such a calculation, the above-described “blur” effect and “calculation” effect are obtained.

[0039]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described below based on an embodiment shown in the drawings. Note that the same or equivalent components as those already described above are denoted by the same reference numerals, and redundant description is omitted.

(1) First Embodiment FIG. 1 is a block diagram showing the configuration of this embodiment.

In FIG. 1, reference numeral 10 denotes an image storage means, which includes a first image memory 11 and a second image memory 12 for storing images of a predetermined area of the strip 1 detected by the detector 3, respectively. ing. Reference numeral 20 denotes a first defective area determination circuit serving as a defective area determination means, which is a binarization circuit 21 for binarizing an output image from the detector 3, and which detects a defect including a plurality of portions from the pattern of the binary image. A crowd determination circuit 22 for determining one connection and a first cutout circuit 23 for determining the separation of the defect region determined to be one connection and determining each defect region are provided.

Reference numeral 30 denotes an identification circuit as identification means for identifying the type and grade of a defect included in the determined defect area. When a defective area is determined by the defective area determination circuit 20, an image of a certain area is output from one of the first and second image memories 11 and 12 to the identification circuit 30, and the other is output from the detector 3. The new image to be output is stored.

Next, referring to FIG.
The operation of determining a defective area by 0 will be described.

The binarizing circuit 21 discriminates the input image at an appropriate judgment level to obtain a binary image (FIG. 2A). The crowd determination circuit 22 determines the “connection” of the parts constituting each defect by performing expansion processing or thickening processing on the pattern of the binary image. That is, by expanding (thickening) the defect to a certain area around the point of the defect, adjacent defects are connected. As a result, FIG.
A, B, C, and D in (b) are a series of defects each composed of a region surrounded by a dotted line.

The first extraction circuit 23 determines the separation of a series of defects determined by the crowd determination circuit 22, and determines each defect area. As the determination process, for example, a labeling process is performed. In this method, the connectivity of each pixel constituting a defect is determined, and each defective area is labeled. This determination method takes a relatively long time because the processing becomes complicated.

On the other hand, as another processing method, it is also possible to use a projection type processing method for determining a defective area from the oblique shadow of the strip 1 in the flow direction or the width direction. This method is relatively simple and fast.

The processing method of the projection system will be described with reference to FIG. In the figure, A, B, C, D is the defect of 1 connections determined in the crowd judging circuit 22, to obtain a P _y which projection to the direction of flow strip 1. This projection cuts, obtaining for example T _1, the projection P _x 1 in the width direction at T _2, P _{x 2.} From the projections in this flow direction and width direction, A, B, C, D
Are obtained, and these areas are determined as the respective defect areas since S _a , S _b , S _c , and S _d are obtained.

As described above, according to this embodiment,
Even if one defect is divided into several parts, and each part has a complicated shape, determination of one connected defect by expansion processing or the like and determination of one connected defect by projection method processing or the like are performed. By performing separation or the like, it is possible to determine a defect area in a simple process at a high speed in almost real time, and to further identify the type and grade of the defect included in the defect area.

(2) Second Embodiment FIG. 4 is a block diagram of the present embodiment.

In FIG. 4, reference numeral 40 denotes a second defective area determining circuit as a defective area determining means, which receives an output image from the detector 3 and "blurrs" a defect having a high density or a defect having a low density. Multi-valued image filter circuits 4 and 1 for performing a process of “accumulating” and floating, a binarizing circuit 22 for binarizing an image subjected to such processing, and a defect area based on a pattern of the binarized image. And a second cutout circuit 43 for determining each defect area by determining the separation. When a defective area is determined by the second defective area determination circuit 40, an image of a certain area is output from the image memory 10 to the identification circuit 30.

FIG. 5 shows the configuration of the multi-value image filter circuit 41. A local image memory 5 for receiving an output image from the detector 3 and storing an image necessary for performing multi-value image processing
1, a filter memory 52 for storing parameters for performing "blur" and "calculation" operations, and a calculation circuit 53 for performing calculations between the image memory output and the filter memory output.

Next, the operation of determining a defective area by the second defective area determination circuit 40 will be described with reference to FIGS.

FIG. 6A is an example of an image detected by the detector 3. E and F are defects. The shaded portions are "dark" portions having a high signal level, and the portions indicated by dots are "light" portions having a low signal level. When this image is binarized as it is, the image becomes as shown in FIG. 2B, and one defect is divided into a plurality of portions. Special processing is required to determine this as one connected defect.

However, when the image shown in FIG. 9A is blurred, that is, when the level of the portion around the high signal level is increased and then binarized, the image becomes as shown in FIG. , E, and F become one connected defect. “Bokeh”
By doing so, the undulation of the signal level is reduced.

FIG. 7A shows another example of an image detected by the detector 3. G, H, and I are one connected defect. Each of the defects has a low signal level as a whole and very few high level portions. If such an image is binarized as it is, as shown in FIG. 4B, only a high-level portion remains slightly dotted, and the original defective portion cannot be reproduced from this binary image. Can not.

Therefore, the image of FIG. 9A is "integrated", that is, the signal level of a certain pixel is calculated by integrating the signal level of the pixel with the integrated value of the pixel contained in an appropriate area around the pixel (hereinafter referred to as a filter window). When the binarization is performed and then the binarization is performed, each of the G, H, and I defective portions becomes one connected portion as shown in FIG. By integrating, the noise component of the background is averaged, but when a defective portion enters the filter window, the level is low but integrated, so that the defective portion "floats".
That's why.

The above-described multi-value image filter circuit 41 gives the above-mentioned "blur" to the input image signal or performs "integration".

The binarization circuit 22 receives the output of the multi-level image filter and discriminates it at an appropriate judgment level to obtain a binary image. As described above, one defect can be obtained as a continuous binary image even if the density of the defect is high as described above or the entire image is thin and blurred.

The second extraction circuit 43 outputs the binary
Each defect area is determined by determining the separation of one connected defect that has been quantified. As the determination process, for example, a labeling process is performed. In this method, the connectivity of each pixel constituting a defect is determined, and each defective area is labeled. This determination method takes a relatively long time because the processing becomes complicated. On the other hand, as another processing method, the strip 1
It is also possible to adopt an oblique processing method of determining a defective area from oblique shadows in the flow direction and width direction. This method is relatively simple and fast.

The operation of the multi-value image filter 41 (see FIG. 4) will be described with reference to FIG. FIG.
FIG. 8A is an explanatory diagram of a filter that gives the above-described “blur” effect, and FIG. 8B is an explanatory diagram of a filter that gives an “integration” effect. All show operation waveforms.

First, FIG. 8A will be described. The signal S indicates one scan of the image signal detected by the detector 3. However, the signal level of the portion corresponding to the background is set to zero. J ₀ parts, K portions are each defect. When discriminating the signal directly at the discrimination level DL0, but binary signal shown in BS0 is obtained, J _0, K each defect, even though a respective one connection, a plurality of pulse signals in the binary signal Has been broken down. This is because only a portion having a high signal level remains as a binarized signal because the density fluctuation of the defective portion is large.

Therefore, the signal FS is obtained by giving the signal S a "blurring" effect. Since the level change of the defective portion of this signal is small and has a smooth waveform,
When binarized by L, it becomes a signal BS, and it can be seen that each defective portion of J ₀ and KA becomes one pulse each.

Next, FIG. 8B will be described. The signals S1, S2, S3,... S10, S11,. Low defect density (thin, pale)
Therefore, the level of the defective portion L only slightly decreases, and cannot be discriminated from the noise level of the background. The signal FS is obtained by moving the filter window and integrating the scanning signals in this window. At the defective portion, the background noise is random in each scan, whereas the signal level has a small variation, but is present in each scan, so that it is taken out as a sufficient variation by integration. By discriminating this FS at the discrimination level DL, the defective portion is extracted as one continuous pulse as shown in the binary signal BS.

The operation and algorithm of the multi-valued image filter 41 will be described in more detail with reference to FIGS.

FIG. 9 shows the relationship between the input image and the filter window. FIG. 3A shows an example of an input image configuration, and one section of the grid corresponds to a pixel. In the case of this figure, 1 in the width direction (I).
7, 18 in the longitudinal direction (J), thus 17 × 18 = 306
Consists of pixels. For example, 0.5m pixels on a 1m x 2m surface
If m × 0.5 mm, (2 × 10 ³ ) × (4 × 1
0 ³ ) = 8 × 10 ⁶ pixels, which is a larger capacity image than FIG.

The pixel (I, J) assumes a value f (I, J) corresponding to the density of the object detected by the detector.
In the figure, the hatched pixels are the I-th in the width direction and the J-direction in the longitudinal direction.
Pixel, which is referred to as pixel (I, J). The region indicated by the thick line surrounding this pixel is the filter window described above. In this example, the shape of the filter window is a square and is composed of 5 × 5 = 25 pixels.

FIG. 9B shows this filter window. The center address is (0, 0), and the value of the i-th pixel (i, j) in the width direction and the j-th pixel (i, j) in the longitudinal direction is g (i, j). Filter window shape, size, and filter function g
(I, j) is determined by the purpose of the filter.

FIGS. 10A to 10D show various filters. Each pixel of the filter is denoted by a symbol such as A or B. For example, the filter function value of the pixel denoted by A is represented by g
(A).

FIG. 9A shows an example of a filter used for the “blur” processing. The filter function value is, for example, g (A) = 1 /
169, g (B) = 3/169, g (C) = 9/169
And the filtering operation, that is, the conversion from the input image f (I, J) to the filtered image F (I, J)

(Equation 2) And _{However, l 1, l 2, m} 1, m 2 is a parameter related to the area of the filter window, in the example of FIG. (A) l ₁ =
l ₂ = m ₁ = m ₂ = 2.

The meaning of the expression (2) will be described below. Pixel (I,
J), the value f (I + i, J) of the neighboring pixel
+ J) and an operation for obtaining the product of the filter function g (i, j) are performed by changing i and j from −l ₁ to l ₂ and −m ₁ to m ₂ , respectively, and adding each multiplied value.

The product sum value F (I, J) obtained in this manner
Is the filtered value of pixel (I, J). I and J perform the above operation by moving all points of the input image. Note that the filter function of this example means that a weight of 1: 3: 9 is applied to the input image toward the periphery with the point of interest A as the center, and the product sum is obtained. All weights 1 /
Sometimes 25 is added.

FIG. 9A can also be used as a filter for the "integration" effect for a planar defect having a low density (thin). For example, g (A) = g (B) = g (C) = 1
age,

(Equation 3) To obtain the converted image. This is equation (2).
g (i, j) = 1.
Equation (3) means the value near pixel (I, J), f (I
+ I, J + j) inside the filter window, i.e., the sum of i and j in the ranges of -l _{1 to} l ₂ and -m _{1 to} m ₂ is defined as the filtered pixel value F (I, J). That is.

For example, among the surface defects of the rolled material, there are defects that are long in the longitudinal direction, such as scratches and slivers. The filter shown in FIG. 10B is effective for extracting low-density (thin) defects. That is, by making the filter window long in the longitudinal direction,
This enhances the "accumulation" effect. In the case of FIG.
For example, f (A) = f (B) = 1 and equation (3) may be calculated. In this example, the vertical line is used. However, in consideration of the width of the defect or the movement of the object, two lines,
Column filters are also good.

There is also a defect that extends thinly in the width direction. For example, a slight indentation created when a roll stops in a rolling process due to a defect called a roll stop.
To catch this, it is effective to use a filter window that is long in the width direction as shown in FIG. The calculation is, for example, f
(A) = f (B) = 1 and the equation (3) may be calculated.

The filter shown in FIG. 6D is the same as that shown in FIG.
(C) in FIG. 3 and is an effective filter for both long defects in the longitudinal direction and long defects in the width direction.

The filters shown in FIGS. 7B, 7C, and 7D can be applied by setting an appropriate filter function even for a linear defect having a large change in density. .

[0077]

As described above, according to the present invention,
Detecting means for detecting an image of a surface defect of the traveling belt-like object; defect area determining means for determining a defect distribution state from the detected image of the surface defect to determine a defect area composed of the same kind of defects; And identification means for identifying the type and grade of the defects included in the defect area, the surface defects of various shapes on the strip, even if the surface defects consist of a plurality of parts, The type and grade of the defect can be quickly and accurately identified.

[0078]

According to the present invention, the defect area determining means includes a binarizing circuit for binarizing the image of the surface defect, and determining the same type of defect from the binarized binary image pattern.
Since it is composed of a crowd judging circuit for judging the connection and a cutout circuit for judging the separation of the defect judged to be one connection and determining each defect area, it surrounds the same kind of defect such as a line or a dot. The defect area can be accurately determined.

[0080]

[0081]

According to the present invention, the cut-out circuit also takes the projections of the defects determined as one connection in the longitudinal direction and the width direction of the strip, and determines the respective defect areas from the projections. With this method, each defect area can be determined more easily and at higher speed than the above-described labeling processing.

According to the present invention, in the above-mentioned projection system, the cutout circuit takes a projection in the width direction at each end of the projection of the defect in the longitudinal direction of the band-like object, and Since each of the defect areas is determined by the rectangular area determined by the projection, the simpler and higher-speed processing can be performed.

[0084]

[0085]

According to the present invention, even when the density difference of each part of one defect is large, even when the density of one defect is entirely low, the area of the defect can be determined accurately.

According to the ninth aspect of the present invention, it is possible to perform "blurring" processing and "integration" processing on the surface defect image. With this processing, a defect area can be accurately determined when binarized.

[0088]

[0089]

[0090]

[0091]

According to the present invention, for a surface defect image,
“Blur” processing and “integration” processing are possible, and by performing this processing, the defect area can be accurately determined.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a first embodiment of the present invention.

FIG. 2 is a diagram illustrating a process of determining a defective area by a defective area determination circuit in the first embodiment.

FIG. 3 is a diagram for explaining a defect area determining operation of a cutout circuit by a projection method in the first embodiment.

FIG. 4 is a block diagram showing a second embodiment of the present invention.

FIG. 5 is a block diagram of a multi-valued image filter circuit according to the second embodiment.

FIG. 6 is a diagram for explaining a defective area determining operation in which a density change is drastic in the defective area determination circuit of the second embodiment.

FIG. 7 is a diagram for explaining an operation of determining a low-density defect area in the defect area determination circuit according to the second embodiment.

FIG. 8 is a waveform chart for explaining the operation of the multi-value image filter.

FIG. 9 is a diagram for explaining the operation of the multi-value image filter circuit.

FIG. 10 is another diagram for explaining the operation of the multi-value image filter circuit.

FIG. 11 is a block diagram of a conventional surface inspection apparatus.

FIG. 12 is a diagram for explaining the operation of the conventional example.

[Explanation of symbols]

 REFERENCE SIGNS LIST 1 strip 3 detector (detection means) 11, 12 first and second image memories 20 first defect area determination circuit (defect area determination means) 21 binarization circuit 22 crowd determination circuit 23 first cutout circuit 30 Identification circuit (identification means) 40 Second defective area determination circuit 41 Multi-valued image filter circuit 43 Second extraction circuit 51 Local image memory 52 Filter memory 53 Arithmetic circuit

Continuation of front page (56) References JP-A-5-280959 (JP, A) JP-A-6-11458 (JP, A) JP-A-6-207909 (JP, A) JP-A-5-180620 (JP, A) , A) JP-A-59-180451 (JP, A) JP-A-55-121584 (JP, A) JP-A-3-124254 (JP, U) (58) Fields investigated (Int. Cl. ⁷ , DB G01N 21/84-21/958 G06T 1/00-9/40

Claims (3)

(57) [Claims] 1. A detecting means for detecting an image of a surface defect of a traveling strip, a binarizing circuit for binarizing the image of the surface defect detected by the detecting means, A crowd judging circuit for judging the same kind of defect as one connection from the pattern of the binary image, and a projection in the width direction at each end of the projection of the defect in the longitudinal direction of the band-shaped object judged to be one connection. A defect region determining means comprising: a first cutout circuit for determining each of the defect regions by a rectangular region determined by the projection in the longitudinal direction and the width direction; and a type and a class of a defect included in the determined defect region. A surface inspection apparatus, comprising: identification means for identifying 2. A detecting means for detecting an image of a surface defect of a traveling belt-like object, and a range for performing multi-value image processing before binarizing the image of the surface defect detected by the detecting means. Multi-valued image processing including a local image memory for storing an image, a filter memory for storing parameters for multi-valued image processing, and an arithmetic circuit for performing a multi-valued image operation between the local image memory and the filter memory A multi-valued image filter circuit,
A binarization circuit for binarizing the image subjected to the multi-valued image processing; and a binarization circuit for judging a defect distribution state from the pattern of the binarized binarized image to determine a defect region composed of the same type of defect. A surface inspection apparatus comprising: a defect region determining unit including two cutout circuits; and an identification unit that identifies a type and a grade of a defect included in the determined defect region. 3. A detecting means for detecting an image of a surface defect of a running belt-like object, and a range for performing multi-value image processing before binarizing the image of the surface defect detected by the detecting means. A local image memory for storing an image, a filter memory for storing parameters for multi-valued image processing, and an output of the local image memory and an output of the filter memory. A multi-valued image filter circuit for performing multi-valued image processing including an operation circuit for performing a calculation, a binarization circuit for binarizing the image subjected to the multi-valued image processing, and a binarized binary image And a second cutout circuit for determining a defect area composed of the same kind of defects by judging a distribution state of the defects from the pattern of the defect area; and determining a type and a class of the defects included in the determined defect area. Identification means to identify Surface inspection apparatus characterized by comprising.