JP5605010B2 - Surface inspection method - Google Patents

Description

  The present invention relates to a surface inspection method, and more specifically, a surface inspection capable of detecting a scratch or the like existing on the surface of an inspection object by image processing (= inspecting whether a scratch or the like exists). It is about the method.

  A product such as a steel material, an iron material, an aluminum material, or the like may be subjected to surface inspection for inspecting whether or not there is a scratch or the like on the surface after manufacturing. As such a product surface inspection method, for example, a method in which an inspector visually detects a scratch or the like, and a method in which image processing is applied to automatically detect a scratch or the like are used.

  A general surface inspection method to which image processing is applied is performed by the following procedure, for example. First, the surface of a product (= inspection object) is imaged to create image data. Next, a pixel having a luminance value higher than a predetermined threshold (or a luminance value lower than the predetermined threshold) is extracted from the pixels included in the created image data. When a pixel having a luminance value (or a luminance value lower than the predetermined threshold value) exists, the pixel is detected as a scratch. Alternatively, a “pixel set” composed of pixels having a luminance value (or lower luminance value) higher than a predetermined threshold is detected, and the size of the “pixel set” or the pixels included in the “pixel set” When the number exceeds a predetermined value, the “pixel set” is detected as a scratch. Thus, generally, it is determined whether or not “scratches” exist based on the luminance of the pixels of the image data. Finally, the quality of the product is determined based on the “scratches” detected in this way.

  By the way, “the unevenness and the pattern which do not need to be regarded as a scratch” may exist on the surface of the product. For example, a product subjected to grinding or cutting may have processing marks such as grinding marks or cutting marks on the surface thereof. Such processing marks may not be regarded as scratches. When the surface inspection method as described above is applied to such a product, it may be difficult to distinguish between “scratches” and “unevenness or pattern that may not be regarded as scratches”. In other words, in the method for determining whether or not the pixel is “scratched” based on the luminance of the pixel of the image data, the pixel corresponding to “scratch” and “unevenness or pattern that does not need to be regarded as a scratch” has the luminance in the same level band. If they have a value, they cannot be distinguished. For this reason, the accuracy of the inspection for “scratches” may be reduced.

  In some cases, a configuration for removing noise included in image data is used in order to improve the accuracy of the inspection of scratches. For example, even in a set of pixels having a luminance value higher than a predetermined threshold, a method of removing a set of pixels having a size smaller than a predetermined size as noise may be used. In addition, a method of removing noise by averaging the luminance value of each pixel of image data for each predetermined region may be used.

  However, these methods are not effective for the products as described above. That is, if a set of pixels having a size smaller than the predetermined size is regarded as a processing mark and removed, scratches having a size smaller than the predetermined size are also removed and cannot be detected. Then, in order to prevent the detection of a flaw from being detected, if a set of pixels having a small size is detected as a flaw, a processing mark is also detected as a flaw. Further, in the configuration in which the luminance values of the pixels are averaged, when the pixels corresponding to the processing marks are removed, even the pixels corresponding to the scratch are removed. And in order to prevent the detection omission of a crack, if the degree of averaging is made low, a processing trace will not be removed but it will be detected as a crack.

  As a configuration for solving such a problem, for example, a configuration as in Patent Document 1 has been proposed. The configuration described in Patent Literature 1 irradiates a product that has been ground in a predetermined direction with light, and selectively reflects components scattered in a direction perpendicular to the grinding direction from the reflected light from the surface of the product. It is detected and the surface property of the product is inspected based on the detection result. And, by utilizing the amount of information of the scattering component having a distribution that spreads widely in the direction perpendicular to the grinding direction, it is possible to reliably detect uneven surface defects that are difficult to distinguish clearly. . According to the description in Patent Document 1, the reflected light from the surface of the product is based on the knowledge that the spread of the scattered component in the direction perpendicular to the grinding direction is larger than the spread in the other direction. .

  However, the configuration described in Patent Document 1 has a problem that it cannot be applied if the grinding direction is unknown or unknown. That is, as described above, since the component scattered in the direction perpendicular to the grinding direction is selectively detected from the reflected light from the surface of the product, the grinding direction needs to be known. Furthermore, the configuration described in Patent Document 1 has a problem that it is difficult to apply to a product whose grinding direction is not constant or a plurality of products whose grinding directions are different from each other.

Japanese Patent Laid-Open No. 9-196862

  In view of the above situation, the problem to be solved by the present invention is to detect the extending direction of the unevenness or pattern when the surface of the inspection object has unevenness or pattern that is not a scratch extending in a predetermined direction. To provide a surface inspection method that can be performed, or even if there are irregularities and patterns that are not scratches extending in a predetermined direction on the surface of the inspection object, Providing a surface inspection method that can be removed, or even if there are irregularities or patterns that are not scratches on the surface of the object to be inspected and the stretching direction of the irregularities or pattern is not constant, Providing a surface inspection method that can suppress or remove the influence, and also the influence of the irregularities and patterns on a plurality of inspection objects with different uneven directions and pattern extending directions. Providing a surface inspection method capable braking or removed, it is.

In order to solve the above problems, the present invention is a method for inspecting a surface of an inspection object in which unevenness or a pattern extending in a predetermined direction is present in the inspection object area, and images the inspection object area of the inspection object A step of generating data, a step of performing a differentiation process on the luminance value of each pixel of the image data along two directions orthogonal to each other, and a value of the two directions orthogonal to each other of the pixels of the image data. Calculating a vector as a component; calculating a distribution of angles from a predetermined reference direction of the calculated vector;
And a step of specifying the extending direction of the unevenness or pattern based on the distribution of the calculated angle of the vector from a predetermined reference direction.

  The “predetermined direction” is not limited to a known specific direction (= a specific single direction or a specific plurality of directions). Further, it may be a single direction or a plurality of directions. And it does not matter whether the “predetermined direction” is known or unknown in the surface inspection method according to the present invention. That is, in carrying out the surface inspection method according to the embodiment of the present invention, the extending direction of the unevenness and pattern need not be known, and may be unknown.

  The “unevenness or pattern extending in a predetermined direction” is an unevenness or pattern that can be rephrased as “unevenness or pattern having directionality in shape”. Specifically, for example, linear (which may be linear or curved) irregularities and patterns are included. In addition, even an unevenness or pattern having a planar spread includes an unevenness or pattern having a “longitudinal direction” (= “longitudinal direction” can be recognized). For example, the shape is a rectangle or an ellipse.

  In the step of calculating the distribution of angles of the calculated vector from a predetermined reference direction, a total value of the vector sizes is calculated for each angle from the predetermined reference direction, and the calculated vector In the stage of specifying the extending direction of the unevenness or pattern based on the direction distribution, the angle at which the calculated peak of the total value appears and the angle having a phase difference of 90 ° is the extending direction of the unevenness or pattern. A specific configuration can be applied. The “angle at which the peak appears” and “stretching direction of unevenness and pattern” may be values having a certain width.

  The present invention further includes a step of performing correction to increase a luminance value of a pixel of a vector having an angle different from an angle at which the peak value appears, and an inspection target region of the inspection target based on the corrected image data. And a stage for inspecting scratches.

  In the step of performing the correction, it is preferable that the filter represented by (Equation 1) is applied to the luminance value of each pixel of the image data.

  The Sobel differential method or Previt's differential method is applied to the differential processing.

  According to the present invention, when there are many irregularities and patterns extending in a predetermined direction on the surface of the inspection object, the extending direction of the irregularities and patterns can be specified. For example, grinding traces of grinders, cutting traces by various cutting tools, polishing traces or grinding traces by various grindstones or abrasive grains, roller traces by rolling rollers, die friction traces in extrusion or drawing, etc. The extending direction of the processing trace to be performed can be specified. Further, not only the processing trace but also a planar pattern extending in a predetermined direction can be detected. For this reason, in image processing for inspecting scratches, it is possible to remove or reduce the influence of a large number of irregularities and patterns extending in a predetermined direction. Therefore, it is possible to prevent a large number of irregularities and patterns extending in a predetermined direction from being recognized as scratches. As a result, it is possible to improve the accuracy of flaw inspection (= accuracy of flaw detection).

  In other words, in the configuration in which detection is performed based only on the luminance value as in the prior art, if “scratches” and “unevenness or pattern that does not have to be regarded as scratches” have similar brightness, they cannot be distinguished. For this reason, “unevenness or pattern that does not need to be regarded as a scratch” may be detected as “scratch”, or “scratch” may be buried in “unevenness that may not be regarded as a scratch” and cannot be detected. On the other hand, according to the surface inspection method according to the embodiment of the present invention, it is possible to distinguish between “scratches” and “unevenness or pattern that does not have to be regarded as scratches”. Therefore, only “scratches” can be detected, and the accuracy of the inspection of the scratches can be improved.

  In the present invention, the extending direction of “unevenness or pattern that may not be regarded as a scratch” may be unknown or unknown. Further, in the stage of imaging the surface of the inspection object, the extending direction of “unevenness or pattern that does not have to be regarded as a scratch” does not matter. That is, there is no need to adjust the imaging means and illumination and “unevenness or pattern that does not have to be regarded as a scratch” in a specific direction. For this reason, the contents of the inspection are not complicated, and the labor for the inspection does not increase. Moreover, it can implement with the apparatus of a simple structure.

  Further, in the configuration described in Patent Document 1, when inspecting a plurality of inspection objects, the directions of the inspection objects (= extending direction of “unevenness or pattern that does not have to be regarded as a scratch”) are aligned. There is a need. On the other hand, in the present invention, even when a plurality of inspection objects are inspected, it is not necessary to align the directions of the inspection objects. For this reason, the work contents in the inspection are not complicated and labor is not increased.

  Further, in the configuration of Patent Document 1, if the extending direction of “unevenness or pattern that does not have to be regarded as a scratch” is not constant, the inspection cannot be performed, or the light source must be regarded as “a scratch. A mechanism is required to follow the extending direction of “good irregularities and patterns”. On the other hand, in the present invention, when the extending direction of “irregularities and patterns that do not have to be regarded as scratches” is not constant, detection is performed by dividing the image into sizes that can be regarded as constant. It becomes possible. Furthermore, it is not necessary to change the light source and the imaging means according to the extending direction of “unevenness that does not have to be regarded as a scratch”. Therefore, the inspection can be performed with a simple configuration. Therefore, it is possible to suppress or prevent an increase in the cost of the device.

It is the figure which showed typically the outline of the structure of the apparatus used in the surface inspection method concerning embodiment of this invention. It is the figure which showed the differential process of image data typically. It is a graph which shows the distribution tendency of the angle of the calculated vector. It is the figure which showed typically the result of the binarization of image data, (a) is the figure to which the surface inspection method concerning embodiment of this invention was applied, (b) is the figure to which the conventional method was applied. It is.

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

  The surface inspection method according to the embodiment of the present invention creates an image data by imaging an inspection target region (= the entire surface of the inspection target or a predetermined part) of the inspection target, and sets the predetermined image data to the predetermined image data. By performing the image processing, it is possible to inspect a flaw existing in the inspection target area of the inspection target (= detect a flaw). In image processing, “indentations and patterns that do not have to be regarded as scratches” that exist in the inspection target area of the inspection object are detected, and the effects of “unevenness and patterns that do not have to be regarded as scratches” are detected from the image data. Remove. For this reason, when inspecting scratches (= detecting scratches) existing in the inspection target region of the inspection target object, it is possible to avoid the influence of “unevenness or pattern that may not be regarded as scratches”. Therefore, it is possible to increase the accuracy of inspection of scratches existing in the inspection target region of the inspection target (= accuracy of detection of scratches). In the surface inspection method according to the embodiment of the present invention, “scratches” refer to “abnormal irregularities and patterns”.

  The inspection object of the surface inspection method according to the embodiment of the present invention is a material or product (for example, various metal materials such as iron material, steel material, aluminum material, etc.) that has been subjected to grinding processing by a grinder or the like. For this reason, many grinding marks exist in the inspection object area of the inspection object. The grinding mark is a linear recess extending in a predetermined direction. In the surface inspection direction according to the embodiment of the present invention, the grinding trace is handled as “unevenness, pattern or pattern that does not have to be regarded as a scratch”. Then, in the scratch inspection, the grinding trace is not detected as a scratch, and irregularities and patterns other than the grinding trace are inspected as a scratch (detected as a scratch). For example, a linear unevenness or pattern extending in a direction different from the extending direction of the grinding mark, or a spotted unevenness or pattern is inspected as a scratch.

  In the surface inspection method according to the embodiment of the present invention, the extending direction of the grinding traces does not have to be specifically clear.

  FIG. 1 is a diagram schematically showing a configuration of an apparatus 1 used in a surface inspection method according to an embodiment of the present invention. As shown in FIG. 1, an apparatus 1 used in a surface inspection method according to an embodiment of the present invention includes an imaging unit 11, an illumination 12, a storage unit 13, a calculation unit 14, and an output unit 15. .

  The imaging unit 11 can image the inspection target area of the inspection target 9 and create image data. Various known digital cameras such as a CCD camera can be applied to the imaging means 11.

  The storage unit 13 can store various data such as image data created by the imaging unit 11 and data created by the calculation unit 14 described later. The computing unit 14 can perform predetermined image processing on the image data created by the imaging unit 11. Furthermore, based on the image processing result, it is possible to inspect the inspection target area for the inspection target 9 for scratches. The output unit 15 can output (for example, display) the results of the image processing and the flaw inspection by the calculation unit 14. The memory | storage means 13, the calculating means 14, and the output means 15 are implement | achieved by various well-known personal computers, workstations, etc., for example.

  The illumination 12 can irradiate the inspection target area of the inspection target 9 with light when the imaging unit 11 images the inspection target area of the inspection target 9. The type and configuration of the illumination 12 are not limited, and various known illumination devices can be applied. However, it is preferable that the illumination 12 is capable of irradiating light with uniform intensity to the inspection target region of the inspection target 9. For example, coaxial epi-illumination can be suitably applied. As shown in FIG. 1, the coaxial epi-illumination includes a light source 121 and a half mirror 122, reflects light emitted from the light source 121 to the half mirror 122, and irradiates the inspection target region of the inspection target 9 with the reflected light. can do. And it arrange | positions so that the optical axis of the reflected light in the half mirror 122 and the optical axis of the imaging means 11 may correspond. In the apparatus 1 used in the surface inspection method according to the embodiment of the present invention, the light source 12 is not essential. For example, in an environment where uniform light is irradiated onto the inspection target region of the inspection target 9, the apparatus 1 used in the surface inspection method according to the embodiment of the present invention may not include the light source 12.

  The surface inspection method according to the embodiment of the present invention includes (1) a step of creating an image data by imaging an inspection target area of an inspection target, and (2) performing a predetermined image processing on the image data to be regarded as “a scratch. (3) a step of specifying the stretching direction of “unevenness or pattern that does not need to be”, (3) a step of correcting the image data to remove the influence of “unevenness or pattern that does not need to be regarded as a scratch”, (4) corrected image Inspecting for scratches based on the data.

  Details of each stage are as follows.

(1) Stage in which image data is created by imaging the inspection target area of the inspection target In this stage, first, light is irradiated to the inspection target area of the inspection target 9 by the illumination 12, and the inspection target is detected by the imaging means 11. 9 areas to be inspected are imaged. The imaging means 11 decomposes the inspection object area of the inspection object into a predetermined number of pixels, and converts the luminance of each pixel (= intensity of reflected light from the inspection object area of the inspection object 9) into a numeric code. Create image data. The storage unit 13 stores the created image data. It should be noted that the direction in which the grinding mark extends may be in any direction during imaging, and there is no need to limit (= adjust) the direction of the inspection object 9. Alternatively, the entire inspection target area of the inspection target 9 may be imaged at a time, and one image data may be created for the entire inspection target area of the inspection target 9, or the divided and captured multiple times. A plurality of image data may be created.

  On the surface of the inspection object 9, there are grinding marks that are concave portions. And the light of the illumination 12 is irregularly reflected in the grinding mark. For this reason, the reflected light from the grinding mark is stronger than the reflected light from the other part (= the part that is not the recess). Further, if there is a scratch on the surface of the inspection object 9, the reflected light from the scratch becomes stronger than the reflected light from other portions for the same reason as the grinding mark. Therefore, among the pixels included in the image data, the pixels that show grinding marks and scratches (= the pixels corresponding to the positions of the grinding marks and scratches) are the pixels that show other parts (= the parts that are not uneven or patterned) The luminance value is higher than that of the pixel corresponding to (1).

(2) A step of performing predetermined image processing on the image data to specify the extending direction of “unevenness or pattern that may not be regarded as a scratch”. This step further includes (2-1) the luminance of each pixel of the image data (2-2) a step of calculating a vector based on the result of the differential processing, (2-3) a “flaw and a defect existing in the inspection target area of the inspection target based on the calculated vector; Identifying the direction of “unevenness or pattern that may not be considered”.

(2-1) Stage of performing a differentiation process on the luminance value of each pixel of the image data At this stage, the calculation means 14 adds the image data stored in the storage means 13 to the X-axis direction and the Y-axis direction. A differential process is performed (= differential processes are performed in two directions orthogonal to each other). That is, the difference in luminance value between pixels of the image data is calculated in two directions, the X-axis direction and the Y-axis direction. For this differentiation processing, for example, various known digital image differentiation methods such as Sobel's differentiation method and Previt's (Prewit) differentiation method can be applied.

  For example, in the Sobel differential method, a Sobel filter (also referred to as “Sobel operator”) shown in (Expression 1-a) and (Expression 1-b) is used for the luminance value of each pixel of image data. Apply. The filter shown in (Formula 1-a) is an example of a Sobel filter for performing a differentiation process in the X-axis direction. The filter shown in (Formula 1-b) is an example of a Sobel filter for performing differentiation processing in the Y-axis direction.

  Further, in the case of the Previt's differential method, the Previt filter shown in (Expression 2-a) and (Expression 2-b) (also referred to as “Previt operator”) with respect to the luminance value of each pixel of the image data. Apply. The filter represented by (Expression 2-a) is an example of a Previt filter for performing differentiation processing along the X-axis direction. The filter shown in (Equation 2-b) is an example of a Previt filter for performing differentiation processing along the Y-axis direction.

  In addition, in Formula (1-a), Formula (1-b), Formula (2-a), and Formula (2-b), a filter having a 3 × 3 element is shown. The number is not particularly limited. Further, the applied filter may be a filter on a square matrix, or may be a filter having a different number of elements in the row direction and the column direction. Further, the differentiation method is not limited to Sobel's differentiation method or Previt's differentiation method, and other various digital image differentiation methods and digital data differentiation methods can be applied.

  Such differentiation processing is performed for all pixels. When this differentiation process is performed, the differential value in the X-axis direction and the differential value in the Y-axis direction of the luminance value are calculated for all pixels included in the image data.

  FIG. 2 is a diagram schematically showing differentiation processing of image data. “Target image data” in FIG. 2 shows an example of image data to be subjected to differentiation processing. Here, three types of target image data are shown as examples. Target image data No. Reference numeral 1 denotes image data including bright lines extending parallel to the reference direction (in the embodiment of the present invention, the X-axis direction is the reference direction). Target image data No. Reference numeral 2 denotes image data including bright lines extending in a direction perpendicular to the reference direction (= Y-axis direction). Target image data No. Reference numeral 3 denotes image data including a bright line extending in the direction of 45 ° with respect to the reference direction and a bright line extending in the direction of 135 °. The “reference direction” is a direction set for convenience in order to indicate the direction of the grinding mark. In the embodiment of the present invention, the X-axis direction is the reference direction, but is not particularly limited. Further, the “bright line” is assumed to be “a collection of pixels having a higher luminance value than the surroundings in a linear form”. The “X-axis direction differential result” and the “Y-axis direction differential result” are the luminance value differential processing result (= differential value) along the X-axis direction of the “target image data” and the Y-axis direction, respectively. The result of the luminance value differentiation process is shown. The pixels having a small differential value are shown dark (= black), and the pixels having a large differential value are shown bright (= white).

  As shown in FIG. When a differential process is performed along the X-axis direction, the luminance value changes rapidly at the end points of the bright lines. For this reason, at the end point of the bright line, the differential value is a predetermined value that is not 0 (= a value corresponding to the difference in luminance value between the bright line and the other part). At other positions, the luminance value does not change between pixels adjacent in the X-axis direction. For this reason, the differential value is 0. On the other hand, when differential processing is performed along the Y-axis direction, the differential value and the extending direction of the bright line are orthogonal to each other, so that the luminance value changes abruptly at the position of the bright line. For this reason, the differential value at the position of the bright line is a predetermined value that is not zero. At other positions, the differential value is zero.

  Therefore, the target image data No. When the differential processing is performed on 1, the pixel corresponding to the bright line has a differential value in the X-axis direction of 0 (zero) except for the pixel corresponding to the end point, and the differential value in the Y-axis direction is a predetermined value that is not 0. It becomes. On the other hand, in the pixels other than the pixel corresponding to the bright line, the differential value in the X-axis direction and the differential value in the Y-axis direction are both 0.

  Target image data No. When differential processing is performed on the X-axis direction 2, the differential direction and the bright line extending direction are orthogonal to each other, so that the luminance value changes abruptly at the position of the bright line. For this reason, at the position of the bright line, the differential value in the X-axis direction is a predetermined value that is not zero. At other positions, the differential value in the X-axis direction is zero. On the other hand, when the differentiation process is performed in the Y-axis direction, the luminance value abruptly changes at the end point of the bright line. Therefore, at the end point of the bright line, the differential value in the Y-axis direction is not a predetermined value (= the bright line and other values). Value corresponding to the difference in luminance value of the portion). At other positions, the luminance value does not change between pixels adjacent in the Y-axis direction, so the differential value in the Y-axis direction is zero.

  Therefore, the target image data No. When the differential processing is applied to 2, the pixel corresponding to the bright line is a non-zero differential value in the X-axis direction and the differential value in the Y-axis direction is 0 except for the pixel corresponding to the end point. On the other hand, for the pixels other than the pixel corresponding to the bright line, the differential value in the X-axis direction and the differential value in the Y-axis direction are both zero.

  Target image data No. 3 shows that when a differential process is performed along the X-axis direction, the differential value and the extending direction of the bright line intersect at a predetermined angle (= 45 °, 135 °), so that the luminance value changes abruptly at the position of the bright line. To do. For this reason, the differential value in the X-axis direction of the pixel corresponding to the bright line is a predetermined value that is not zero. The differential value in the X-axis direction of pixels other than the pixel corresponding to the bright line is zero. The same applies when the differential processing is performed in the Y-axis direction.

  Therefore, the target image data No. When the differential process is applied to 3, the differential value of the pixel corresponding to the bright line becomes a predetermined value that is not 0 in both the X-axis direction and the Y-axis direction. In addition, the specific magnitude | size of a differential value changes according to the intersection angle of a bright line and a differential direction. For example, when the crossing angle between the bright line and the differential direction becomes smaller, the differential value becomes smaller and becomes larger as it approaches 90 °. The target image data No. 3, since the angle of the bright line with respect to the differential direction is the same in the X-axis direction and the Y-axis direction, the differential value in the X-axis direction and the differential value in the Y-axis direction of the pixel corresponding to the bright line are equal.

(2-2) Stage of calculating a vector based on the result of the differentiation process In this stage, for each pixel included in the image data, a vector V having a differential value in the X-axis direction and a differential value in the Y-axis direction as components. Is calculated. Furthermore, the magnitude | V | of the calculated vector V of each pixel and the angle θ with respect to the reference direction (here, the X-axis direction) are calculated. The magnitude | V | of the vector V of each pixel and the angle θ with respect to the reference direction are

| V | = ((differential value in the X-axis direction) ² + (differential value in the Y-axis direction) ² ) ^1/2
θ = tan ⁻¹ ((differential value in the Y-axis direction) / (differential value in the X-axis direction))

Is calculated by

  The calculated angle θ of the vector V of each pixel indicates the direction in which the change in the luminance value between the pixel and the surrounding pixels is the largest. For this reason, the direction perpendicular to the angle θ of the vector V (= the direction having a phase difference of 90 °) is the direction in which the change in luminance between each pixel and the surrounding pixels is the smallest. Also, the magnitude | V | of the vector V indicates the rate of change in luminance between the pixel and the surrounding pixels. For this reason, a pixel with a large differential value has a large difference in luminance value from the surroundings.

(2-3) Step of identifying the direction of “unevenness or pattern that does not need to be regarded as a scratch” existing in the inspection target region of the inspection target based on the calculated vector. Create a graph like this: The graph shown in FIG. 3 is a graph showing the distribution of angles of the calculated vector V weighted by the magnitude of the vector V. And the tendency of the distribution of the angle which the vector V has is shown. The horizontal axis of the graph shown in FIG. 3 is the angle θ with respect to the reference direction of the vector V of each pixel. The vertical axis represents the total value of the vectors V at each angle. More specifically, the image data, when the pixel of the vector V with a predetermined angle theta _a with respect to the X axis is included the N is the is at a predetermined angle theta _a, of the N pixels The total value of the magnitudes | V | of the vector V is plotted.

  Target image data No. In 1, the differential value in the X-axis direction of the pixel corresponding to the bright line is 0 (zero), and the differential value in the Y-axis direction is a predetermined value that is not 0. For this reason, the direction of the vector V of the pixel corresponding to the bright line has a phase difference of 90 ° (= 270 °) with respect to the X axis. On the other hand, the differential value in the X-axis direction and the differential value in the Y-axis direction of pixels other than the pixel corresponding to the bright line are both zero. That is, the vector V of pixels other than the pixel corresponding to the bright line is a so-called zero vector. Therefore, the target image data No. In the graph of 1, the total value of the magnitudes of the vectors V protrudes at the positions of 90 ° and 270 ° (= peak appears), and becomes 0 at other angles.

  Target image data No. In 2, the differential value in the X-axis direction of the pixel corresponding to the bright line is a predetermined value that is not 0, and the differential value in the Y-axis direction is 0. For this reason, the direction of the vector V of the pixel corresponding to the bright line is a direction parallel to the X axis (the phase difference from the X axis is 0 ° (= 180 °)). On the other hand, for the pixels other than the pixel corresponding to the bright line, the differential value in the X-axis direction and the differential value in the Y-axis direction are both zero. For this reason, the vector V of pixels other than the pixel corresponding to the bright line is a so-called zero vector. Therefore, the target image data No. In the graph of 2, the total value of the magnitudes of the vectors V protrudes at the positions of 0 ° and 180 °, and becomes 0 at other angles.

  Target image data No. 3, the differential value of the luminance of the pixel corresponding to the bright line having an angle of 45 ° with respect to the reference direction is a predetermined value that is not 0 in both the X-axis direction and the Y-axis direction. Since the intersection angle between the bright line and the X and Y axes is 45 ° (= 225 °), the differential value is the same in the X axis direction and the Y axis direction. Therefore, since the vector V of the pixel corresponding to this bright line has the same absolute value in the X-axis direction component and the Y-axis direction component, the direction has a phase difference of 135 ° (= 315 °) with respect to the reference direction. Have. On the other hand, the pixel vector V corresponding to the portion other than the bright line is a so-called zero vector. The same applies to pixels corresponding to bright lines having an angle of 45 ° with respect to the reference direction.

  Therefore, the target image data No. In the graph of 3, the total value of the magnitudes of the vector V protrudes at angles of 45 °, 135 °, 225 °, and 315 °, and becomes 0 at other angles.

  As described above, when the pixel having the vector V having the same angle with respect to the reference direction includes a certain number of pixels in the image data, the total value of the sizes of the vector V at the angle is obtained at other angles. The value is larger than the total value. That is, the peak of the total value of the vector V appears at the angle. Further, when pixels having vectors V having different angles with respect to the reference direction are included in the image data, the total value of the sizes of the vectors V at the respective angles is obtained at other angles. It becomes larger than the total value. That is, at each angle, a peak having the magnitude of the vector V appears (= a plurality of peaks appear in the graph of FIG. 3).

  The angle of each vector V has a phase difference of 90 ° with respect to the extending direction of the bright line. For this reason, the angle having a phase difference of 90 ° with respect to the angle at which the peak of the total value appears is the bright line drawing direction. As described above, when the image data including the bright line is subjected to the processes of the steps (2-1), (2-2), and (2-3), the extending direction of the bright line can be specified. it can.

  The grinding marks are substantially parallel to each other and extend in a specific direction. For this reason, among the pixels included in the image data, the vector V of the pixel corresponding to the grinding mark (= the pixel in which the grinding mark appears) has a substantially constant angle with respect to the reference direction. Further, the grinding mark is brighter than the other parts because the illumination light is irregularly reflected. For this reason, the brightness | luminance of the pixel corresponding to a grinding mark becomes higher than a pixel other than that. Therefore, the grinding marks are included as bright lines in the image data. And since the number of grinding marks is larger than that of scratches and other irregularities and patterns, the number of pixels corresponding to the grinding marks is the number of pixels corresponding to other parts (scratches and other irregularities and patterns). More than.

  For this reason, when the graph shown in FIG. 3 is created for the image data obtained by imaging the inspection target region of the inspection target, a peak of the total value appears at an angle having a phase difference of 90 ° with respect to the extending direction of the grinding mark. . Therefore, an angle having a phase difference of 90 ° at an angle at which the peak of the total value appears can be specified as the extending direction of the grinding mark.

  As described above, when the image data obtained by imaging the inspection target area of the inspection target is subjected to the processes of the steps (2-1), (2-2), and (2-3), the grinding marks are stretched. The direction can be specified. In actual processing, it is only necessary to have a data set that can create the graph shown in FIG. 3, and it is not necessary to actually create the graph shown in FIG. In short, it suffices if the calculation means 14 can determine at which angle the calculated total value of the vector V has a peak.

(3) A step of correcting the image data to remove the influence of “unevenness or pattern that does not have to be regarded as scratches” In this step, the image data is subjected to predetermined image processing, and pixels included in the image data are corrected. Correct the brightness value. Then, the influence of the pixel corresponding to the grinding mark is removed (or the influence of the pixel corresponding to the grinding mark is reduced). Specifically, a filter that emphasizes the luminance value along the direction perpendicular to the extending direction of the grinding mark is applied to the image data.

  This filter is a filter in which a circle is modeled by a trigonometric function. That is, in the 3 × 3 matrix, 8 elements excluding the center (element b-2) indicate the respective positions on the circumference. Specifically, (element b-3) indicates a position of 0 ° with respect to the X axis, (element a-3) indicates a position of 45 ° with respect to the X axis, and (element a-2) indicates A 90 ° position with respect to the X axis, (element a-1) indicating a position of 135 ° with respect to the X axis, (element b-1) indicating a position of 180 ° with respect to the X axis, (Element c-1) indicates a position of 225 ° with respect to the X axis, (Element c-2) indicates a position of 270 ° with respect to the X axis, and (Element c-3) indicates a position with respect to the X axis. A position of 315 ° is shown.

  When this filter is applied to the luminance value of each pixel of the image data, the luminance at the angle (= direction) assigned to “ξ” of each element of this filter is enhanced. That is, the (element b-2) of this filter is overlaid on the pixel to be applied (= target pixel). Then, the sum of the product of the pixel of interest and each of the eight pixels adjacent to the pixel of interest and the value of the element overlapping each of these pixels is replaced with the luminance value of the pixel of interest (= new luminance value) . That is,

  {(Element a-1) × (Luminance value of upper left pixel of target pixel) + (Element a-2) × (Luminance value of pixel above target pixel) + (Element a-3) × (Target pixel (Brightness value of upper right pixel) + (element b-1) × (brightness value of pixel to the left of the target pixel) + (element b-2) × (brightness value of the target pixel) + (element b-3) × ( The luminance value of the pixel to the right of the pixel of interest) + (element c-1) × (the luminance value of the pixel at the lower left of the pixel of interest) + (element c-2) × (the luminance value of the pixel below the pixel of interest) + ( Element c-3) × (luminance value of lower right pixel of target pixel)} / 9

  And the result is set as a new luminance value of the pixel of interest. In the above formula, “right” is the positive direction of the X axis, “left” is the negative direction of the X axis, “up” is the positive direction of the Y axis, and “down” is the negative direction of the Y axis. It is.

  When such a calculation is performed, the luminance value of the pixel of interest is the luminance of the pixel adjacent to the angle (= angle from the reference direction) assigned to the element “ξ” among the luminance values of the pixels adjacent to the pixel of interest. Is replaced with the emphasized luminance value. When the pixel of interest is a pixel corresponding to the bright line, a pixel adjacent in a direction perpendicular to the extending direction of the bright line of the pixel is a pixel having a low luminance value. For this reason, if an angle perpendicular to the extending direction of the bright line is substituted for “ξ” of the filter element, the luminance value of the pixel of interest is reduced by enhancing the luminance value of these pixels (= corrected to a low value). ) Since the luminance value of the pixel adjacent in the bright line extending direction is high, if the angle parallel to the bright line extending direction is substituted for “ξ” of the filter element, the luminance value of the pixel of interest is emphasized to a high value. Replaced.

  Therefore, when the extending direction of the grinding mark has a phase difference of 125 ° with respect to the reference direction, 215 ° or 35 ° which is an angle perpendicular to 125 ° is set to “ξ” of the filter element. substitute. Then, the luminance value of the pixel corresponding to the bright line is corrected to a low value.

  At this stage, the luminance value of the pixel corresponding to the grinding mark is corrected to a low value, or the luminance value of the pixel other than the pixel corresponding to the grinding mark is corrected to a high value. Good. For this reason, it is not limited to the structure which applies the said filter. For example, a configuration in which the luminance value of a pixel having a vector V in a direction parallel to the extending direction of the grinding trace is reduced, or a configuration in which the luminance value of a pixel having a vector V in a direction different from the extending direction of the grinding trace is increased is used. Applicable. Moreover, the structure which correct | amends the luminance value of each pixel of image data according to the phase difference of the extending | stretching direction of a grinding trace and the vector V may be sufficient.

(4) Stage of Inspecting Scratches Based on Corrected Image Data In this stage, first, the calculation means 14 binarizes the luminance value of each pixel of the image data corrected in the stage (3). That is, a pixel having a luminance value equal to or higher than a predetermined threshold (or a pixel exceeding the predetermined threshold) is set as a “white pixel”, and a pixel having a luminance value lower than the predetermined threshold (or a pixel equal to or lower than the predetermined threshold) is set as a “black pixel”. By the correction in the step (3), the luminance value of the pixel corresponding to the grinding mark is lower than the luminance value of the pixel corresponding to the scratch. Then, a binarization threshold value is set between the luminance value of the pixel corresponding to the grinding mark and the luminance value of the pixel corresponding to the scratch. Then, in the binarized image data, the pixel corresponding to the scratch is a “white pixel”, and the other pixels (including the pixel corresponding to the grinding mark) are “black pixels”.

  Note that the binarization threshold is appropriately set according to the degree of correction in the step (3).

  Then, based on the binarized image data, the inspection object area is inspected for scratches. A conventionally known method can be applied to the scratch inspection method. For example, when “a set of white pixels” is detected from binarized image data and the number of white pixels included in one “set of white pixels” is greater than a predetermined number, the detected “white pixels” A method of regarding “a set of” as “scratches” can be applied. Also, the size of the detected “collection of white pixels” (X-axis direction dimension or Y-axis direction dimension) is calculated, and if the size exceeds a predetermined dimension, the detected “white pixel set” is defined as “scratch”. Applicable methods are applicable. Thus, the method for inspecting scratches based on the binarized image data is not particularly limited.

  FIG. 4 is a diagram schematically showing binarized image data. Then, (a) shows a case where the surface inspection method according to the embodiment of the present invention is applied, and (b) shows a case where a conventional method (= method without performing the correction in the step (3)) is applied. It is a thing. In the conventional method, if the pixel corresponding to the scratch and the pixel corresponding to the grinding mark have the same level of luminance value, it is difficult to distinguish them. For example, when the threshold value for binarization is lowered in order to prevent detection of scratches from being detected, the pixel corresponding to the grinding mark becomes a “white pixel”. Then, grinding marks are detected as “scratches”. On the other hand, if the binarization threshold is increased in order to prevent the grinding trace from being detected as a scratch, the pixel corresponding to the “scratch” to be detected becomes the “black pixel”. As a result, scratches to be detected cannot be detected. Thus, in the conventional method, if there is a grinding mark in the inspection target region of the inspection target, it is impossible to accurately inspect the scratch.

  On the other hand, in the surface inspection method according to the embodiment of the present invention, in the image data immediately after being imaged, the pixel corresponding to the scratch and the grinding mark (that is, “irregularity that does not have to be regarded as a scratch or Even if the pixels corresponding to the “pattern”) have similar luminance values, the luminance values of these pixels can be differentiated by subsequent image processing. That is, the luminance value of the pixels other than the pixel corresponding to the grinding mark can be corrected to a high value without affecting the luminance value of the pixel corresponding to the grinding mark. Alternatively, the luminance value of the pixel corresponding to the grinding mark can be corrected to a low value without affecting the luminance value of the pixel corresponding to the scratch.

  Then, when binarizing the corrected image data, setting a binarization threshold value between the luminance value of the pixel corresponding to the grinding mark and the luminance value of the pixel corresponding to the scratch corresponds to the scratch. Only the pixels to be processed can be made “white pixels” (= the pixels corresponding to the grinding marks can be made not to be “white pixels”). For this reason, it is possible not to detect pixels corresponding to the grinding marks as scratches. In this way, it is possible to distinguish between a pixel corresponding to a scratch and a pixel corresponding to a grinding mark (= a pixel corresponding to “an unevenness or pattern that does not need to be regarded as a scratch”). That is, the influence of grinding marks can be removed from the image data. Therefore, it is possible to improve the accuracy of inspection of scratches (improving the accuracy of detection of scratches).

  The “angle at which the peak appears” and “the extending direction of the grinding mark” may be values having a certain width. That is, the configuration may be such that the extending direction of the grinding mark is specified based on an angle corresponding to “one point with a mountain (for example, the peak of the mountain of the graph)” in the graph shown in FIG. The configuration may be such that the extending direction of the grinding mark is specified based on a certain range of angles before and after the “mountain apex”. Therefore, in the step (3), when applying the filter to the luminance value of each pixel of the image data, one angle in the “ξ” of the filter is substituted and the filter is applied only once. It is also possible to employ a configuration in which a certain range of angles is substituted and the filter is applied a plurality of times.

  In addition, even when a plurality of types of grinding traces having different stretching directions (= different angles with respect to the reference direction) are mixed in the inspection target region of the inspection target, all of the plurality of types of grinding traces are mixed. The influence can be removed. That is, in such a case, the target image data No. 1 in FIG. As shown in the graph 3, a peak of the total value of the vector V appears at each of a plurality of angles corresponding to the extending direction of the grinding mark. Then, the step (3) is applied to each angle at which the peak of the total value of the vector V appears. If it does so, all the influences of the grinding trace from which a extending | stretching direction mutually differs can be removed.

  For example, the target image data No. 3, when a peak of the total value appears at 45 °, 135 °, 225 °, and 315 °, a grinding mark having an angle of 45 ° (= 225 °) with respect to the reference direction, and 135 ° ( = 315 °), it can be determined that there are two types of grinding marks having an angle. In such a case, in the step (3), 0 ° (or 180 °) is substituted for ξ of the filter shown in the equation (3) and applied to the image data. Further, 90 ° (or 270 °) is substituted for ξ and applied to the image data. That is, when there are a plurality of extending directions of the grinding mark, an angle value that does not correspond to any of these angles is substituted into the filter shown in Expression (3) and applied. For example, as described above, an intermediate angle between grinding traces extending in different directions is substituted. With such a configuration, it is possible to increase the luminance value of pixels other than the pixel corresponding to the grinding mark. In other words, the luminance value of the pixel corresponding to the grinding mark can be lowered.

  When the direction of the grinding process by the grinder is constant over the entire inspection target area of the inspection object 9, the extending direction of the grinding trace is also constant over the entire inspection target area of the inspection object 9. For this reason, in this case, the entire inspection target region of the inspection target 9 can be imaged to create one piece of image data, and the entire piece of the one piece of image data can be processed collectively. However, the inspection target area of the inspection target 9 may be divided into a plurality of parts depending on the performance of the personal computer as the computing means 14 and the like, and processing may be performed for each of the divided areas. In this case, since the size of the image data to be processed can be reduced, it is possible to prevent the calculation means 14 from being overloaded.

  On the other hand, when the direction of the grinding process by the grinder is not constant (= different depending on the location), the extending direction of the grinding mark is also not constant (= different depending on the location). In this case, the following method can be applied.

  First, image data created by imaging an inspection target area of an inspection target is divided to generate a plurality of image data. Alternatively, the inspection target area of the inspection target is divided and images are taken for each of the divided areas to create a plurality of image data. Then, each of the plurality of image data is subjected to the processing of each stage. If the direction of the grinding process by the grinder is not constant, the extending direction of the grinding mark is not constant as a whole. However, even in such a case, it can be considered that the extending direction of the grinding mark is substantially constant locally. Accordingly, a plurality of pieces of image data having a size such that the extending direction of the grinding mark can be regarded as uniform is created from the region subjected to the grinding process. Then, processing is performed for each created image data. Thereby, the extending direction of the grinding mark can be specified in each of the plurality of image data. Then, using each of the plurality of image data, it is possible to remove the influence of the grinding mark and inspect the scratch. For this reason, the inspection precision of a crack can be aimed at.

  Note that, as described above, the size of the divided image data and the range to be imaged are sizes that allow the extending direction of the grinding marks to be considered constant. Therefore, the size of the divided image data and the range to be imaged are appropriately set according to the shape and dimensions of the grinding mark (for example, the radius of curvature).

  In addition, when it is unknown whether the direction of the grinding mark is constant, when the shape of the grinding mark is unknown, or when the size and shape of the grinding mark are not constant (for example, when the radius of curvature varies depending on the location) For example, the following method can be applied.

  First, the entire inspection target area of the inspection target 9 is imaged to create one piece of image data. Then, the processes of steps (2) and (3) are performed on the created image data. If the direction of the grinding mark cannot be specified even after performing the steps (2) and (3), the image data created first is divided into a plurality of small-size image data. Then, the processes of the steps (2) and (3) are performed on the divided image data. Alternatively, a part of the range is extracted from the first created image data, and the processing in the steps (2) and (3) is performed on the extracted range.

  If the extending direction of the grinding mark cannot be specified even after dividing the image data, the image data is further divided to create image data of a smaller size. If the extending direction of the grinding mark cannot be specified even if a part of the range is extracted from the image data, image data in which the extracted range is further reduced is created. The image data created in this way is subjected to the processes of steps (2) and (3).

  Such an operation is repeated until the extending direction of the grinding mark can be specified. As described above, in the steps (2) and (3), the method of gradually reducing the size of the image data to be processed (that is, the range in which scratches are inspected in one process) is applied. it can. If it is such a method, even if it is a case where the extending direction of a grinding trace is unknown, or the shape of a grinding trace is unknown, the extending direction of a grinding trace can be specified.

  Note that “the extending direction of the grinding mark cannot be specified” means that when the graph shown in FIG. 3 is created, the peak of the total value of the vector V does not appear at a specific angle (in other words, it is dominant. In the case where there is no angle having a total value protruding so as to allow a difference). For example, this is the case when the total value is substantially equal at all angles.

  Note that the size of division or the size of extraction differs depending on the shape of the grinding mark (for example, the radius of curvature). For this reason, it can test | inspect by appropriate size by reducing size in steps. In other words, if the size of the divided image data or the size of the extracted region is a size that allows the extending direction of the grinding marks to be considered to be substantially constant, the influence of the grinding marks can be removed from the image data. become.

  The effects of the surface inspection method according to the embodiment of the present invention are summarized as follows.

  When there are a large number of irregularities and patterns extending in a predetermined direction in the inspection target area of the inspection object, the extending direction of the irregularities and patterns can be specified. For example, the extending direction of a processing mark extending in a predetermined direction such as a grinding mark of a grinder can be specified. For this reason, in image processing for inspecting scratches, it is possible to remove or reduce the influence of a large number of irregularities and patterns extending in a predetermined direction. Therefore, it is possible to prevent a large number of irregularities and patterns extending in such a predetermined direction from being recognized as scratches. As a result, it is possible to improve the accuracy of flaw inspection (= accuracy of flaw detection).

  In other words, in the conventional configuration in which the detection is based only on the brightness, if “scratches” and “unevenness or pattern that does not have to be regarded as scratches” have the same brightness, they should be distinguished. I can't. For this reason, “unevenness or pattern that does not need to be regarded as a scratch” may be detected as “scratch”, or “scratch” may be buried in “unevenness that may not be regarded as a scratch” and cannot be detected. On the other hand, according to the surface inspection method according to the embodiment of the present invention, it is possible to distinguish between “scratches” and “unevenness or pattern that does not have to be regarded as scratches”. For this reason, “scratches” can emerge from “unevenness or pattern that does not have to be regarded as scratches”. Further, “irregularities and patterns that do not have to be regarded as scratches” can be removed. As a result, only “scratches” can be detected.

  Further, in the surface inspection method according to the embodiment of the present invention, the extending direction of “unevenness or pattern that may not be regarded as a scratch” may be unknown. In the stage where the inspection target area of the inspection target is imaged and image data is created, the extending direction of the “unevenness or pattern that does not have to be regarded as a scratch” is not a problem. For this reason, the content of inspection does not become complicated and labor is not increased. Moreover, the apparatus used in the surface inspection method according to the embodiment of the present invention does not need to be complicated.

  For example, in the surface inspection method according to the embodiment of the present invention, the extending direction of “unevenness or pattern that does not have to be regarded as a scratch” is specified in the image processing stage of the inspection. For this reason, the “unevenness or pattern that does not have to be regarded as scratches” stretching direction need not be apparent in the previous stage. Further, there is no need to have a specific relationship between the illumination or the imaging means and the direction of “unevenness or pattern that may not be regarded as a scratch”. Further, in the surface inspection method according to the embodiment of the present invention, even when a plurality of inspection objects are inspected, it is not necessary to align the directions of the inspection objects. For this reason, the work contents in the inspection are not complicated, and the labor of the work does not increase.

  Further, in the surface inspection method according to the embodiment of the present invention, when the extending direction of “unevenness that may not be regarded as scratches” is not constant, the image is divided to such an extent that it can be regarded as constant. , Detection becomes possible. Furthermore, it is not necessary to change the light source and the imaging means according to the extending direction of “unevenness that does not have to be regarded as a scratch”. Therefore, the inspection can be performed with a simple configuration. Therefore, it is possible to suppress or prevent an increase in the cost of the device.

  The embodiments of the present invention have been described in detail with reference to the drawings. However, the present invention is not limited to the embodiments, and various modifications can be made without departing from the spirit of the present invention. Is possible.

  For example, in the above-described embodiment, grinding marks by a grinder are shown as “unevenness or pattern that does not have to be regarded as scratches”, but the present invention is not limited to this. The present invention can also be applied to various types of processing marks such as cutting marks by various cutting tools, polishing marks and grinding marks by various grinding stones and abrasive grains, roller marks by rolling rollers, and die friction marks in extrusion and drawing processes. Furthermore, the present invention can be applied not only to processing traces but also to predetermined irregularities and patterns formed on the surface of a product (= inspection object).

  That is, in the surface inspection method according to the embodiment of the present invention, the “unevenness or pattern that does not need to be regarded as a scratch” may be “unevenness or pattern having directionality in shape”. In addition, a line-shaped (which may be a straight line shape or a curved line shape) unevenness or pattern is included. In addition, even an unevenness or pattern having a planar spread includes an unevenness or pattern having a “longitudinal direction” (= “longitudinal direction” can be recognized). For example, the shape is a rectangle or an ellipse.

  In addition, as described above, the “predetermined direction” of the “unevenness or pattern that may not be regarded as a scratch” is limited to a known specific direction (= a specific single direction or specific directions). It is not something. And it does not matter whether the “predetermined direction” is known or unknown in the surface inspection method according to the present invention. That is, in carrying out the surface inspection method according to the embodiment of the present invention, the extending direction of the unevenness and pattern need not be known, and may be unknown. Furthermore, it may be a single direction or a plurality of directions.

  Further, in the surface inspection method according to the embodiment of the present invention, one “unevenness or pattern that may not be regarded as a scratch” does not have to be in a predetermined single direction. For example, one “unevenness or pattern that does not have to be regarded as a scratch” may be linear or curved. In short, when a part of the inspection target area is extracted, each “unevenness or pattern that does not have to be regarded as a scratch” in the extracted area may be in a certain predetermined direction. (As long as it has directionality in the extracted region). In addition, a plurality of “unevenness and patterns that do not have to be regarded as scratches” existing in the extracted area may be oriented in a certain predetermined direction (= with directionality in the extracted area) If any).

  Furthermore, in the above-described embodiment, a pixel having a luminance higher than the predetermined threshold is regarded as a pixel corresponding to “scratch” or “a concave portion (unevenness or pattern) that may not be regarded as a scratch”. The configuration is not limited. For example, a configuration may be adopted in which a pixel having a luminance lower than a predetermined threshold is regarded as a pixel corresponding to a “scratch” or “a concave portion (unevenness or pattern) that may not be regarded as a scratch”. The point is that the pixels corresponding to “scratches” and “recesses (irregularities and patterns) that do not have to be regarded as scratches” are appropriately set according to whether the luminance is higher or lower than the other pixels. Good.

  The imaging unit 11 may be a so-called area scan camera or a line scan camera. When an area scan camera is applied to the image pickup means 11, it is possible to pick up an image of the entire inspection target area of the inspection object or a predetermined part and having a two-dimensional extent at a time. When a line scan camera is applied to the imaging unit 11, image data having a two-dimensional spread can be created by capturing an image while relatively moving the imaging unit and the inspection target.

  Furthermore, in the said embodiment, although the grinding trace (namely, recessed part) was shown as "an unevenness | corrugation and a pattern which do not need to be regarded as a crack", it is applicable even if it is a planar pattern. In short, any image recognition can be applied as long as it is recognizable (for example, if the brightness value and the color are different from other parts).

DESCRIPTION OF SYMBOLS 1 Apparatus used in the surface inspection method concerning embodiment of this invention 11 Imaging means 12 Illumination 121 Light source 122 Half mirror 13 Storage means 14 Calculation means 15 Output means 9 Inspection object

Claims (5)

A method for inspecting the surface of an inspection object having unevenness and patterns extending in a predetermined direction in the inspection object region,
Imaging the inspection target area of the inspection target and creating image data;
Performing a differentiation process on the luminance value of each pixel of the image data along two directions orthogonal to each other;
Calculating a vector whose component is a differential value in two directions orthogonal to each other for each pixel of the image data;
A step of calculating the angle size of the sum of the vector for each of the predetermined reference direction, to calculate the angular distribution from the predetermined reference direction of the calculated said vector,
Identifying the extending direction of the unevenness or pattern based on the distribution of angles from the predetermined reference direction of the calculated vector;
Performing correction to increase the luminance value of a vector pixel having an angle different from the angle at which the peak value of the total value appears, and removing or reducing the influence of the unevenness and pattern from the image data;
Inspecting the inspection object for scratches on the inspection object based on the corrected image data;
A surface inspection method comprising: In the step of identifying the extending direction of the convex-concave or pattern based on the direction of the distribution of the vector pre SL calculated, the angle having a phase difference of angle and 90 ° for peak appears the calculated the sum the uneven Ya surface inspection method according to claim 1, wherein the benzalkonium be identified as the extending direction of the pattern. The correction in the step of performing is a surface inspection method according to claim 1 or 2, characterized by applying the filter shown in the luminance value of each pixel of the image data (Equation 1).

The surface inspection method according to claim 3, wherein an angle perpendicular to the extending direction of the unevenness or pattern is substituted for “ξ” in the formula (1). 5. The surface inspection method according to claim 1, wherein a Sobel differential method or a Previtt differential method is applied to the differential processing. 6.

Images