JP2001126064A - Method for detecting defect on metallic surface - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve inspection without depending on the amount of chamfering, to reliably detect a small defect such as a fine flaw and to improve inspection without depending on the roughness of a surface in a method for detecting a defect on a metallic surface, which detects the defect on the surface of a matter to be inspected by photographing the matter, with a TV camera 24 and giving picture processing to this photographic picture. SOLUTION: An inspection area on the surface of a matter to be inspected 22 is divided into plural fine areas (step 3), a standard deviation concerning luminance is measured in each of these fine areas (step 5), the degree of correlation between this standard deviation and a standard deviation in a master work free from a defect, which is measured and registered in advance, is calculated (step 12) and this degree of correlation is compared with a threshold for detecting a defect, which is set in advance (step 14) to detect the presence/ absence of the defect on the surface of the matter 22.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for detecting a defect on a metal surface, and more particularly, to a method for detecting a defect on a metal surface having a chamfered portion typified by an end face of a bearing.

[0002]

2. Description of the Related Art In recent parts production lines, the number of inspections of defects on the surface of a workpiece by image processing technology at the end of a processing step has increased. Specifically, the image processing apparatus has at least an image processing device, illumination, and a TV camera, captures illumination light reflected on the surface of the workpiece by the TV camera, and converts the captured image into various images by the image processing device. The processing technology is applied to determine the presence or absence of a defect. As an image processing technique in this case, a template matching method and a method of extracting an image feature amount such as an area are generally used.

[0003] In the inner and outer rings of bearings (inner and outer rings), their raceways (surfaces in contact with rolling elements such as balls and rollers)
And end face defect inspection. Here, considering the inner and outer ring end surfaces of the bearing, as shown in FIG. 3, the respective boundary portions between the outer peripheral surface and the inner peripheral surface of the end surface are chamfered. This chamfer was made in the chamfering process in the bearing production line,
In general, chamfering does not greatly affect the processing accuracy, so that chamfering is often performed relatively roughly. Therefore, the size of each of the end face portion and the chamfered portion may vary considerably depending on the workpiece.

[0004]

Here, the problem when the template matching method most frequently used as the image processing technique is applied to the defect inspection of the inner and outer race end faces of the bearing will be described. First, as mentioned above,
Since the inner and outer ring end faces of the bearing are chamfered and the tolerance of this chamfer amount is large, the size of the end face part to be inspected and the size of the chamfered part vary considerably depending on the workpiece. As a result, good inspection cannot be performed by the template matching method on the assumption that the size of the inspection part is constant. Second, in the template matching method, it is difficult to detect a small defect such as a minute flaw because the acquired image data of the workpiece and the image data of the template are compared with each other. Third, the surface roughness of the processed end surface may vary depending on the workpiece depending on the type of the grinding wheel that has processed the end surface, the degree of wear thereof, and the like. In this case, depending on the setting of the threshold value for defect determination, In some cases, those having poor surface roughness may be erroneously recognized as defects.

Further, in the case where the end face has a mark, the following problem occurs in addition to the above problem. First, since the inspection of the lack of the mark and the inspection of the defect except the mark must be performed, the phase of the mark must first be matched. In this case, the phase matching with the master pattern is performed. Since template matching is performed later, the inspection time becomes very long especially when the image size is increased to increase the detection accuracy. Second, since the size and length of the stamp in the acquired image changes depending on the depth of the stamp, this may be erroneously recognized as a defect. Third, in the defect inspection by the template matching method, it is necessary to mask a certain range including the stamp as a dead zone, so that a defect near the stamp cannot be detected.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-mentioned problems, and it is an object of the present invention to be able to perform a good inspection irrespective of the amount of chamfering and to surely remove small defects such as minute scratches. That is, it can be detected, and good inspection can be performed without depending on the surface roughness. In addition, in addition to the above-described objects, shortening the inspection time including the phase alignment of the mark, performing a good inspection independently of the depth of the mark, Another object is to detect defects.

[0007]

According to the first aspect of the present invention, an object to be inspected is photographed by a TV camera, and the photographed image is subjected to image processing. In the defect detection method of the metal surface so as to detect the defect of the surface of the inspection object, the inspection region of the surface of the inspection object is divided into a plurality of minute regions, and a standard deviation related to luminance is measured in each of the minute regions, By calculating the degree of correlation between this standard deviation and the standard deviation of a master work having no defect that has been measured and registered in advance, and comparing this degree of correlation with a preset threshold value for defect detection, And a method for detecting a defect on a metal surface, wherein the method detects the presence or absence of a defect on the surface of the inspection object.

According to the first aspect of the present invention, the standard deviation of the brightness in the corresponding area between the master work and the inspection object is compared. May not necessarily be equal. Further, if the size of the divided area is reduced by using the standard deviation, detection of a small defect such as a minute flaw can be easily dealt with. Furthermore, since the defect can be detected in consideration of the surface roughness by using the standard deviation, it is not possible to erroneously determine that a defect exists only due to poor surface roughness.

In order to achieve the above-mentioned objects, the invention according to claim 2 is the invention according to claim 1, wherein the standard deviation for each minute area of the surface of the inspection object is set in advance. If there is a standard deviation equal to or higher than the level of the engraved mark recognition, the defect of the engraved mark is also inspected. The specific processing procedure is as follows. First, the amount of displacement is calculated from the degree of correlation between the standard deviation of each minute area on the surface of the object to be inspected and the standard deviation of each minute area on the master work. After the alignment of the data of the workpiece and the master work, the part having the standard deviation equal to or higher than the marking recognition level for each of the inspection object and the master work is recognized as a marking block. The difference between the degree of correlation and the width of the engraved block is obtained for each corresponding engraved block, and the degree of correlation is smaller than a preset threshold value, or the difference in the engraved block width is a preset allowable value. If either of them is larger than the above, it is determined that a defect exists in the engraved portion.

Here, the operation of the invention according to claim 2 will be described. Since the standard deviation regarding the luminance is greatly different between a place where the marking is present and a place where the marking is not present, first, in the defect inspection of the marking, data alignment between the surface of the inspection object and the master work is performed. Next, a portion having a standard deviation equal to or higher than the marking recognition level for each of the inspection object and the master work is recognized as a marking block. Then, a correlation between the standard deviation of each minute region on the surface of the inspection object and the standard deviation of each minute region of the master work is obtained for each of the corresponding engraved blocks of the inspection object and the master work, and this correlation is set in advance. If the threshold value is smaller than the threshold value, it is determined that a defect exists in the engraved portion. Further, a difference in the width of the engraved block is determined for each of the corresponding engraved blocks of the inspection object and the master work, and when the difference is larger than a preset allowable value, the defect in the engraved portion is also determined. Is determined.

According to the second aspect of the present invention, the data alignment between the surface of the inspection object and the master work is performed based on the correlation between the standard deviations of the two luminances. There is no need to set a large size, and therefore, the time for positioning the mark is shorter than in the template matching method. According to the second aspect of the present invention, a portion having a standard deviation equal to or higher than the marking recognition level for each of the inspection object and the master work is recognized as a marking block. In order to detect the mark defects based on the correlation between the standard deviation of each minute area on the surface of the inspection object and the standard deviation of each minute area on the master work and the difference in the width of the mark block, template matching is performed. A good inspection can be performed without depending on the depth of the marking as in the method. further,
According to the second aspect of the present invention, for the defect inspection of the inspection area other than the marking, the data obtained by canceling (deleting) the data of the standard deviation of the part recognized as the marking block from the data of the standard deviation of the entire inspection area is used. I just need to do
Unlike the template matching method, it is not necessary to mask a certain range including the inspection area near the marking, and therefore, the defect can be detected also in the inspection area near the marking.

[0012]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS One embodiment of the present invention will be described below with reference to the drawings. FIG. 2 shows an example of a system configuration including an image processing apparatus to which the defect inspection method of the present invention is applied. The bearing 22 as an object to be inspected is placed on the inspection table 21 such that the end surface as the inspection surface faces upward. The upper part of the bearing 22 has C
A TV camera 24 such as a CD camera is provided, and its optical axis is perpendicular to the end face of the bearing 22 as an inspection surface. A ring illumination 23 as an illuminating device is provided between the inspection table 21 and the TV camera 24, and the optical axis of the TV camera 24 passes through substantially the center of an annular portion of the ring illumination 23. The reason for adopting such a configuration is that illumination is uniformly applied to the annular inspection surface,
This is for efficiently receiving the reflected light.

An image processing device 25 obtains a photographed image captured by the TV camera 24 and performs image processing described later. The image processing device 25 has a storage device for storing image data such as a captured image and a master image. The panel key 26 is used by the operator to operate the image processing apparatus 2.
5 is used to input data, command codes, and the like. The monitor TV 27 displays various images such as a photographed image and an image-processed image, and is also used for displaying various setting parameters and inspection results. In addition, if necessary, a loader for automatically carrying in and out the bearing 22 as the inspection object onto the inspection table 21 may be provided.

Here, the processing performed in the image processing device 25 will be briefly described with reference to FIG. 12 showing a control circuit in the image processing device 25. When an operation command is input from the outside, the CPU 37 as a control processor
Signals are sent to an address generation circuit 34 for controlling generation of an address to the AM 33 and a synchronization control circuit 32 for controlling a synchronization signal of the TV camera 24. The TV camera 24 receives a signal from the synchronization control circuit 32, captures an image of the inspection object, and transmits the captured image to the A / D converter 31. A / D
The converter 31 converts the input video signal into 8-bit 256
Is converted into gradation data, and this is converted into VR as an image memory.
Store it in AM33.

Next, the CPU 37 sets the designated area in the address generation circuit 34 and sets the contents of the calculation in the image operation circuit 35, so that the data in the designated area is automatically transferred from the VRAM 33 to the image operation circuit 35. Is done. When the designated operation is executed in the image operation circuit 35 and the operation is completed, the address generation circuit 34 transmits an end signal to the CPU 37, and the CPU 37 receives the operation result from the image operation circuit 35. The D / A converter 36 converts the above-mentioned 8-bit data of 256 gradations into analog data for image display, the RAM 38 is used as a memory for storing registered data such as non-defective data and the like, and a ROM 39 is used as a control memory. It is a memory for storing programs, and the I / O 40 is an interface with an external device.

Hereinafter, the bearing 22 according to the present embodiment will be described.
Will be described with reference to the flowchart shown in FIG. 1 and FIGS.

Step 1: Input an image. In this step, an image is input from the TV camera 24. The image captured by the TV camera 24 is analog data, which is converted into digital data of 256 gradations by A / D (analog-digital) conversion processing and stored. The defect inspection is also performed for the inner ring of the bearing 22 in the same procedure as for the outer ring, but only the outer ring will be described here.

Step 2: An edge position is detected. In this step, a boundary (edge) between the end surface and the outer peripheral chamfer and a boundary (edge) between the end surface and the inner peripheral chamfer are set for the outer ring of the bearing 22. In the outer ring of the bearing 22, each boundary between the outer peripheral surface and the inner peripheral surface at the end surface is chamfered, and these are referred to as an outer peripheral chamfer and an inner peripheral chamfer, respectively. In the present embodiment, the inspection is performed in each of the end face, the outer peripheral chamfer, and the inner peripheral chamfer. Accordingly, a process of dividing the captured image captured by the TV camera 24 into respective areas of the end face, the outer peripheral chamfer, and the inner peripheral chamfer is performed first. The specific procedure for this process is
First, as shown in FIG. 3B, the boundary (edge) between the end face and the outer peripheral chamfer with respect to the outer ring of the non-defective bearing 22.
And the boundary (edge) between the end face and the inner peripheral chamfer are set respectively. Then, this data is compared with the image data captured by the TV camera 24, and the most similar part is set as an edge detection point. This detection point is set as a boundary (edge position) of each area. The data of the non-defective bearing 22 is measured in advance for a master work having no defect, and this data is registered in the RAM 38.

Step 3: An inspection area is set. In this step, as shown in FIG. 4, based on the edge position obtained in step 2 and the outer and inner diameters of the outer ring of the bearing 22, three areas, that is, chamfered areas (outer peripheral chamfered portions) A and the inner peripheral chamfered portion C) and the other area (end face B).

Step 4: Check the width of the end face B. In this step, the width of the end face B is calculated based on the end face B set in step 3, and if the width is out of the allowable range, it is determined that the chamfering is defective.
The process proceeds to the determination NG process of No. 6. This is because when the chamfer amount is large, the width of the end face portion B becomes small, while when the chamfer amount is small, the width of the end face portion B becomes large. Note that the width of the end face portion B is determined in Step 2
Can be easily obtained based on the edge position of the boundary between the end face and the outer peripheral chamfer and the edge position of the boundary between the end face and the inner peripheral chamfer.

Step 5: The standard deviation of the entire circumference is measured. In this step, if it is not determined in step 4 that the chamfering is defective, each of the inspection areas (the outer peripheral chamfer A, the end face B, and the inner peripheral chamfer C) is divided into minute areas. The standard deviation in each minute area is measured for the entire circumference of the end face of. Specifically, a minute area is set for each of the three divided areas as shown by a hatched area in FIG. 4, and the minute area is rotated and moved once by one along the annular inspection area while the light The variation (standard deviation) of the reflection intensity of the light is measured.

FIG. 5 shows an example of measurement data in each inspection area. (A), (b), and (c) are measurement data of the outer peripheral chamfered portion A, the end surface portion B, and the inner peripheral chamfered portion C, respectively, where the horizontal axis is the rotational angle position and the vertical axis is the vertical axis. Is the standard deviation. 6 to 8 and 10 to 11 described later, the horizontal axis represents the rotational angle position, and the vertical axis represents the standard deviation. (B) In the figure, the standard deviation at the position where the inscription (the character of “NACHI”) exists is high. In the present embodiment, the measurement data of the end face portion B is used for the defect inspection of the stamp and the defect inspection other than the stamp, while the measurement data of the outer peripheral chamfered portion A and the inner peripheral chamfered portion C are used for undulations and scratches. Used for inspection. FIG. 6A shows an example of measurement data in the case where undulation is present in the inspection area, and FIG. 6B shows an example of measurement data in the case where a flaw is present in the inspection area.

Step 6: It is determined whether or not the end face B is engraved. In this step, as shown in FIG. 7B, if the standard deviation in the minute area of the end face B measured in step 5 has a standard deviation equal to or higher than the preset marking recognition level, the end face B Is determined to exist, and the process proceeds to step 7 and subsequent steps for inspecting for defects in the stamp. On the other hand, if there is no standard deviation equal to or greater than the marking recognition level at the end face B, it is determined that there is no marking at the end face B, and the processing of steps 7 to 11 is skipped, and the process proceeds to step 12. In addition, as for the outer peripheral chamfered portion A and the inner peripheral chamfered portion C, the inspection area does not have any engraving as a matter of course.
Steps 11 to 11 are skipped, and the process proceeds to step 12.

Step 7: The position of the marking is adjusted. In this step, if it is determined in step 6 that the end face B is engraved, the non-defective data (FIG. 7A) and the measurement data (FIG. b)) and alignment. This is because, in order to inspect the defect of the stamp, it is necessary to first match the position (phase) of the stamp portion of the non-defective product data and the measurement data. For details of this alignment, the correlation degree between the two data is obtained by normalized correlation while shifting the measurement data by one degree with respect to the non-defective data centering on the center position of the outer ring (center position of the ring). The amount of phase shift between the two is calculated based on the position where the degree of correlation is the largest, and the position is adjusted by correcting the measurement data position by the amount of phase shift.

Step 8: Check the engraved part. In this step, each of the portions that are equal to or higher than the above-described stamp recognition level set in advance as shown in FIG. 7B is referred to as a stamp block, and a master work having no defect is measured in advance for each stamp block. The engraved block width is obtained from the registered degree of correlation between the non-defective data and the measured data, and the start point and end point of each engraved block, and the difference between the non-defective data and the measured data is obtained. Then, by comparing the difference between the correlation degree and the width with a preset allowable value, the defect determination of the engraved portion is performed.

Step 9: Whether the difference is within an allowable range. In this step, either the degree of correlation between the non-defective data and the measured data in step 8 is smaller than a preset threshold value, or the difference between the engraved block widths is larger than a preset allowable value. In the case of (1), it is determined that a defect exists in the engraved portion, and the masking process of the next step 10 is not performed on the engraved block determined to have the defect. On the other hand, if the degree of correlation between the non-defective data and the measurement data is larger than the preset threshold value and the difference between the engraved block widths is smaller than the preset allowable value, a defect exists in the engraved part. It is determined not to be performed, and the process proceeds to step 10.

Step 10: Mask the engraved portion. In this step, when it is determined in step 9 that there is no defect in the engraved part, masking is performed on the part recognized as the engraved part. FIG. 9 shows an example of masking. This is because it is necessary to mask the data of the engraved portion in the inspection other than the engraved portion thereafter. FIG. 8C shows the masking of the engraved portion by canceling the engraved portion from the measured data in FIG. 8B. FIGS. 8E and 8F show an example in which it is determined that the engraved portion has a defect.

Step 11: The standard deviation is measured again. In this step, the standard deviation for each minute region is measured again for the inspection region after the marking portion is masked in step 10.

Steps 12 to 14: Calculate the degree of correlation and determine the presence or absence of a defect. In these steps, step 5 or step 1
The presence / absence of a defect is determined for the standard deviation data of each minute area measured in step 1 and the standard deviation data of the outer peripheral chamfered portion A and the inner peripheral chamfered portion C jumped from step 6. More specifically, as shown in FIG. 10, the non-defective product data and the measurement data are set for each set comparison range with reference to the correlation level calculation reference level as a level preset based on the allowable surface roughness. Is calculated (step 12), and this is calculated for the entire comparison range (step 1).
3) The lowest correlation degree in the entire comparison range is set as the overall correlation degree, and this is compared with a preset threshold value for defect detection (step 14). If the value falls below the threshold value, it is determined that a defect exists in the inspection area, and the process proceeds to the determination NG process in step 16, while if the overall correlation degree does not fall below the threshold value, the inspection area falls within the inspection area. It is determined that there is no defect in
The process proceeds to the determination OK process of No. Note that it can be determined that a defect exists in a comparison range having a degree of correlation lower than the defect detection threshold. As described above, since the standard deviation regarding the luminance in the corresponding region between the master work and the inspection object is compared, the sizes of the comparison target regions are not necessarily equal to each other as in the template matching method. Good.

FIG. 10 shows a case where the entire circumference of the end face is divided into 12 as a comparison range, but a small defect can be detected by setting a large number of divisions. This is because the comparison range is narrowed by setting the number of divisions to be large, so that defects in the comparison range are relatively emphasized. As a result, even if a small defect is not overlooked, the correlation of the comparison range can be reduced. This is due to the fact that the degree is reduced. In the case of the image having a rough surface, the standard deviation of the luminance measured as shown in FIG. 11B was smaller than that of the image shown in FIG. Since the overall size is large, it can be determined that the surface roughness is defective.

Steps 15 and 16: OK and N
G. In these steps, judgment OK or judgment NG is output based on the result of the above-described comparison processing. These outputs are displayed on the monitor TV 27 or output from the image processing device 2.
5 to an external device via the I / O 40.

As described above, in this embodiment, the preset marking recognition level, correlation calculation reference level, and threshold value for defect detection can be set arbitrarily for each area of the chamfered part and the end face part. , And the range in which the correlation is calculated can be arbitrarily set, so that the defect inspection can be performed with high accuracy. The time required for the above series of processing is
Although it depends on the number of divisions of the entire circumference of the bearing end face, it takes about 0.5 seconds for both the inner and outer rings, so that defect inspection can be performed in a shorter time and with higher accuracy than the conventional template matching method.

[0033]

According to the first aspect of the present invention, there is provided a metal which is configured to detect a defect on the surface of the inspection object by photographing the inspection object by a TV camera and performing image processing on the photographed image. In the surface defect detection method, the inspection area on the surface of the inspection object is divided into a plurality of minute areas, and a standard deviation relating to luminance is measured in each of the minute areas, and the standard deviation is measured and registered in advance. Calculate the degree of correlation with the standard deviation in the master work having no defect, and detect the presence / absence of a defect on the surface of the inspection object by comparing the degree of correlation with a preset threshold value for defect detection. I did it. Thus, since the standard deviation of the luminance in the corresponding region between the master work and the inspection object is compared, the sizes of the comparison target regions may not necessarily be the same as in the template matching method. As a result, even if the object to be inspected is chamfered, good inspection can be performed regardless of the amount of chamfering. Also, by using the standard deviation, if the size of the region to be divided is reduced, a small defect such as a small flaw can be reliably detected.
Furthermore, by using the standard deviation, it is possible to detect a defect in consideration of the surface roughness, so that it is not erroneously determined that a defect exists only due to poor surface roughness. Good inspection can be performed.

According to the second aspect of the present invention, the first aspect is provided.
In the invention according to the present invention, both alignment is performed from the degree of correlation between the standard deviation of each minute region on the surface of the inspection object and the standard deviation of each minute region in the master work, and the standard deviation equal to or higher than the stamp recognition level is obtained for each of the two. The point having the mark is recognized as an engraved block, a difference between the correlation degree and the width of the engraved block is obtained for each of the corresponding engraved blocks, and a defect of the engraved portion is detected based on the difference. Therefore, the inspection time including the phase alignment of the mark can be reduced, good inspection can be performed regardless of the depth of the mark, and a defect near the mark can be detected.

[Brief description of the drawings]

FIG. 1 is an image processing apparatus 2 according to an embodiment of the present invention.
6 is a flowchart illustrating a procedure of a defect inspection performed in Step 5.

FIG. 2 is a diagram showing an example of a system configuration including an image processing device 25 to which the defect inspection method of the present invention is applied.

FIG. 3 is a diagram showing an end surface of a bearing as an inspection object 22, showing an edge detection point when specifying an inspection area.

FIG. 4 is a diagram showing an inspection area of a bearing outer ring as an inspection object in one embodiment of the present invention.

FIG. 5 is a diagram illustrating an example of measurement data in each inspection area according to the embodiment of the present invention.

FIG. 6 is a diagram showing an example of measurement data in a case where swells and scratches exist in an inspection area according to an embodiment of the present invention.

FIG. 7 is a diagram illustrating an example of measurement data, non-defective data, and determination data when a mark is present in an inspection area according to an embodiment of the present invention.

FIG. 8 is a diagram illustrating an example different from FIG. 7 among the measurement data, the non-defective data, and the determination data in a case where a mark is present in the inspection area according to the embodiment of the present invention.

FIG. 9 is a diagram illustrating an example of masking of an engraved portion according to an embodiment of the present invention.

FIG. 10 is a diagram illustrating an inspection method in an inspection area where no stamp is present, according to an embodiment of the present invention.

FIG. 11 is a diagram showing a difference in measurement data due to a difference in surface roughness in one embodiment of the present invention.

FIG. 12 is a block diagram illustrating an internal configuration of an image processing device 25 according to an embodiment of the present invention.

[Explanation of symbols]

 Reference Signs List 21 inspection table 22 bearing outer ring (inspected object) 23 ring illumination (illumination device) 24 TV camera 25 image processing device 26 panel key 27 monitor TV 31 A / D converter 32 synchronization control circuit 33 VRAM 34 address generation circuit 35 image calculation circuit 36 D / A converter 37 CPU 38 RAM 39 ROM 40 I / O

 ──────────────────────────────────────────────────続 き Continued on the front page (72) Inventor Mitsuru Ueda 1-1-1 Fujikoshi Honcho, Toyama City, Toyama Prefecture Fujikoshi Co., Ltd. (72) Inventor Yukihiro Okamoto 1192-8 Kamijima, Namerikawa-shi, Toyama Techno System Co., Ltd. F term (reference) 2G051 AA90 AB07 BA20 CA03 CA04 CB01 DA01 DA06 EA12 EA14 EA20 EB01 EB02 EC03 EC06 ED01 FA10 5B057 AA17 BA02 CA08 CA12 CA16 CB08 CB12 CB16 CD02 CD08 CH08 DA07 DB02 DB09 DC22 DC34

Claims (2)

[Claims] 1. A method for detecting a defect on a surface of a metal object, wherein the object to be inspected is photographed by a TV camera and image processing is performed on the photographed image. The inspection area on the surface is divided into a plurality of minute areas, and a standard deviation relating to luminance is measured in each of the minute areas. The standard deviation is compared with the standard deviation of a master work having no defect which has been measured and registered in advance. Calculating the degree of correlation and comparing the degree of correlation with a preset threshold value for defect detection to detect the presence or absence of a defect on the surface of the inspection object; Defect detection method. 2. If there is a standard deviation equal to or greater than a preset marking recognition level among the standard deviations for each minute area on the surface of the inspection object, the marking defect is also inspected. The processing procedure for the defect inspection is performed by calculating a positional deviation amount from a degree of correlation between the standard deviation of each minute region on the surface of the inspection object and the standard deviation of each minute region in the master work. After the alignment, a point having a standard deviation equal to or more than the marking recognition level for each of the inspection object and the master work is recognized as a marking block, and the correlation degree and the marking are marked for each corresponding marking block of the inspection object and the master work. The difference between the block widths is obtained, and the degree of correlation is smaller than a preset threshold value, or the difference between the stamp block widths is set in advance. For greater than it had been allowed value of any defect detection method of a metal surface according to claim 1, characterized in that so as to determine that there is a defect in the marking unit.