Surface defect detection method and surface defect detection device

Abstract

To provide a surface defect detection method that can more accurately detect a surface defect of a strip-like body.SOLUTION: A surface defect detection method for optically detecting a surface defect of a strip-like body includes: an image acquisition step of detecting reflection light from a strip-like body obtained by illuminating a surface of the strip-like body, imaging a plurality of images including the surface of the strip-like body while scanning the surface thereof relatively, and acquiring a plurality of images including the surface of the strip-like body; an average image calculation step of calculating an average image of the plurality of acquired images; an image correction step of performing a shading correction of each of the plurality of acquired images using the average image to obtain a correction image; and a defect detection step of detecting a surface defect of the strip-like body on the basis of the correction image. The average image calculation step includes, in each of the plurality of images, recognizing an inspection object area where the strip-like body in the image exists, and contributing to the average image only a pixel in the inspection object area.SELECTED DRAWING: Figure 2

Description

[Technical field]

[0001] The present disclosure relates to a surface defect detection method and a surface defect detection device for detecting surface defects on a strip-shaped body, such as a steel plate. [Background technology]

[0002] In recent years, in the manufacturing process of steel products, it is required to detect surface defects of steel materials in hot or cold conditions in order to suppress mass nonconformity and improve yield. The steel materials described in this disclosure include steel products such as seamless steel pipes, welded steel pipes, hot-rolled steel sheets, cold-rolled steel sheets, and thick plates, as well as shaped steel, and semi-finished products such as slabs produced in the process of manufacturing these steel products. As surface inspection methods for optically detecting surface defects of steel materials, the techniques described in Patent Documents 1 to 3, for example, have been conventionally known.

[0003] For example, Patent Document 1 discloses a surface defect detection method in which two distinguishable light sources are used to irradiate illumination light from different directions onto the same inspection target portion of a steel material, images are obtained based on the reflected light of each illumination light, and differential processing is performed between the obtained images to detect surface defects in the inspection target portion.

[0004] As a general technique for optically detecting surface defects of steel materials, including but not limited to the conventional technique described in Patent Document 1, a method is widely used in which the surface of the steel material to be inspected is illuminated, the reflected light is detected by a camera having a two-dimensional or one-dimensional light receiving element to capture an image, and the obtained image is processed to detect defects. A common method for detecting defects in an image is to compare the brightness of each pixel of the image with a predetermined threshold value to recognize whether each pixel is a healthy part or a defective part.

[0005] The following two points are characteristics of inspection for surface defects on steel products.

[0006] First, since many steel products are long, the entire steel product must be scanned and imaged repeatedly in order to inspect the entire steel product. To achieve such scanning, a typical configuration is to fix lighting and a camera to the production line and repeatedly image the steel product being transported through the production line.

[0007] Secondly, the size of the manufactured steel products, for example the product width, varies, so the area in which the steel products are captured in the image captured by the fixed camera changes. Areas in the image in which the steel products are not captured do not need to be inspected, and at the same time, it is necessary to prevent the erroneous detection of defects in such areas. Therefore, it is necessary to dynamically recognize the range in which the steel products are captured in the captured image, i.e., the inspection target area, and to detect defects only within the inspection target area.

[0008] In many cases, raw images captured by a camera have "shading," where the brightness level varies depending on the position in the image, due to effects such as uneven lighting and reduced brightness in the periphery of the camera's field of view due to lens characteristics. For this reason, it has been difficult to accurately recognize healthy and defective areas by simply comparing the brightness value of a captured image with a threshold value. For this reason, preprocessing, known as shading correction, is performed on raw images to correct the brightness unevenness and make the brightness level of healthy areas uniform.

[0009] More specifically, the correction is performed by dividing the captured image by the shading pattern or by subtracting the shading pattern from the captured image using a waveform or an image that indicates the shading pattern. Shading patterns are roughly classified into fixed patterns and dynamic patterns.

[0010] In the case of a fixed pattern, the same pattern is always used for shading correction. The pattern is determined in advance, for example, experimentally. For example, there is a method in which an average image obtained from multiple images taken of defect-free steel material is used as the shading pattern.

[0011] On the other hand, in the case of a dynamic pattern, shading correction is performed on the captured image while calculating or updating the shading pattern sequentially from one or more images captured during inspection. By calculating or updating this pattern in accordance with the change in shading of the captured image, it is possible to improve the uniformity after correction compared to a fixed pattern.

[0012] A basic method of shading correction using such a dynamic pattern is to use a blurred image obtained by performing moving average processing or low-pass filtering on a captured one-dimensional or two-dimensional image as a shading pattern to correct the original image. There is also a method of obtaining a shading pattern by approximating a captured image using a polynomial or the like. Furthermore, improved methods of these are disclosed in Patent Documents 2 and 3.

[0013] For example, Patent Document 2 discloses a shading correction method for a surface inspection device in which a threshold signal is generated based on a first shading correction signal generated from a pre-stored scanning detection signal up to the previous scan, the signal value for the position of the defective part contained in the scanning signal for the current scan is replaced with the first shading correction signal to obtain a corrected scanning signal in real time, and the scanning detection signal for the current scan is normalized.

[0014] For example, Patent Document 3 discloses a method of detecting the image area of ​​the object to be inspected by detecting the boundary between the object to be inspected and the area outside the object to be inspected from an image from a camera, extracting shading components as an image of the same size as the image from the camera by applying a low-pass filter to the image from the camera, approximating the luminance change of each line in the image area of ​​the object to be inspected by a curve, or approximating the luminance change of the entire image area of ​​the object to be inspected by a curved surface, reconstructing the image into one of the same size with the luminance value of the approximated curve or approximated surface, and dividing or subtracting the image from the camera by the reconstructed image to correct shading. [Prior art documents] [Patent documents]

[0015] [Patent Document 1] JP 2016-224069 A [Patent Document 2] JP 2012-112729 A [Patent Document 3] JP 2004-226347 A Summary of the Invention [Problem to be solved by the invention]

[0016] However, the above-mentioned conventional shading correction methods each have problems.

[0017] For example, in a correction method using a fixed pattern, it is difficult to perform sufficient correction under conditions in which shading varies greatly, and it is difficult to improve detection accuracy.

[0018] For example, in the method of determining the shading pattern by moving average processing or low-pass filter processing of the captured image, there is a problem that "folding" occurs around the defect in the corrected image. Also, in the case of a defect with a large area, the signal in the center of the defect is weakened due to the correction effect, resulting in a "hollow" image. Either problem causes changes in the size and area of ​​the defect, which may worsen the accuracy of the judgment at the stage of judging the type and degree of harmfulness of the defect after detection.

[0019] In the method described in Patent Document 2, the current scan signal is corrected by a first shading correction signal based on the previous scan signal. Therefore, when the edge position of the target material shown in the image moves due to meandering or width change of the test target material, the current scan signal is corrected by the first shading correction signal in the area where the edge position has changed, which causes a problem that it is difficult to perform normal shading correction near the edge.

[0020] In the method described in Patent Document 3, if the approximation polynomial assumed in the process of generating a shading pattern by approximating the curve or surface of the original image is not appropriate, the shading of the original image and the shading pattern will diverge. This causes a problem in that the luminance of healthy areas cannot be made uniform.

[0021] The present disclosure has been made in consideration of the above problems, and aims to provide a surface defect detection method and a surface defect detection device capable of detecting surface defects on a strip-shaped body with higher accuracy. For example, the objective is to provide a surface defect detection method and a surface defect detection device that suppresses "folding" around a defect or "hollowing" of a defect with a large area, and also performs appropriate shading correction up to the vicinity of the edge even when the edge position shown in the image fluctuates due to meandering or width change of the inspected material. [Means for solving the problem]

[0022] The present disclosure relates to (1) 1. A method for optically detecting surface defects on a strip, comprising: an image acquisition step of detecting reflected light from the strip obtained by illuminating the surface of the strip, and capturing an image of the surface of the strip while relatively scanning the surface of the strip to acquire a plurality of images including the surface of the strip; an average image calculation step of calculating an average image of the plurality of acquired images; an image correction step of performing shading correction on each of the acquired images using the average image to obtain a corrected image; a defect detection step of detecting surface defects of the strip based on the corrected image; Including, the step of calculating the average image includes identifying an inspection target area in each of the plurality of images in which the band exists, and allowing only pixels in the inspection target area to contribute to the average image. Surface defect detection method, It is.

[0023] (2) In the surface defect detection method described in (1) above, The image correction step may include dividing each of the acquired images by the average image.

[0024] (3) In the surface defect detection method described in (1) above, The image correction step may include subtracting the average image from each of the plurality of acquired images.

[0025] (4) In the surface defect detection method according to any one of (1) to (3), The average image calculation step may include recognizing, in each of the plurality of images, non-constant parts within the inspection region in the image, and allowing only pixels of the constant parts to contribute to the average image.

[0026] (5) In the surface defect detection method according to any one of (1) to (4), The strip may comprise steel.

[0027] The present disclosure relates to (6) A surface defect detection device for optically detecting surface defects on a strip, comprising: an illumination unit that illuminates a surface of the strip; an imaging unit that detects reflected light from the strip obtained by illuminating a surface of the strip with the illumination unit; an image processing unit that processes a plurality of images including the surface of the strip captured by relatively scanning the surface of the strip using the illumination unit and the imaging unit to detect surface defects of the strip; Equipped with The image processing unit includes: Calculating an average image of the plurality of images; obtaining a corrected image by performing shading correction on the plurality of images using the average image; detecting surface defects on the strip based on the corrected image; identifying a region of interest in each of the plurality of images in which the band exists, and contributing to the average image only those pixels that are in the region of interest; Surface defect detection device, It is. Effect of the Invention

[0028] According to the surface defect detection method and the surface defect detection device of an embodiment of the present disclosure, it is possible to detect surface defects of a strip-shaped body with higher accuracy. For example, "folding" around the defect or "hollowing" of a defect with a large area is suppressed. In addition, even if the edge position shown in the image changes due to meandering or width change of the inspected material, it is possible to perform shading correction appropriately up to the vicinity of the edge. [Brief description of the drawings]

[0029] [Figure 1] FIG. 1 is a schematic diagram illustrating a configuration of a surface defect inspection device according to an embodiment of the present disclosure. [Diagram 2] 4 is a flowchart showing image processing executed by the image processing unit in FIG. 1. [Diagram 3] FIG. 2 shows a number of images of the steel material of FIG. [Figure 4] FIG. 1 illustrates an embodiment of the present disclosure. [Diagram 5] FIG. 1 is a diagram showing a comparative example according to the prior art. [Figure 6] FIG. 13 is a diagram showing a comparative example according to another conventional technique. [Figure 7A] FIG. 5 is a graph showing a corrected luminance waveform between A and A in the shading-corrected image of FIG. 4(d). [Figure 7B] FIG. 6 is a graph showing a corrected luminance waveform between A and A in the shading-corrected image of FIG. 5(d). [Figure 7C] FIG. 7 is a graph showing a corrected luminance waveform between U and U in the shading-corrected image of FIG. 6(d). DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0030] Hereinafter, the configuration and operation of a surface defect inspection device according to an embodiment of the present disclosure will be mainly described with reference to the accompanying drawings.

[0031] 1 is a schematic diagram showing the configuration of a surface defect detection device 1 according to an embodiment of the present disclosure. The surface defect detection device 1 optically detects surface defects on a strip. In the present disclosure, the "strip" includes, for example, steel materials. The steel materials described in the present disclosure include steel products such as seamless steel pipes, welded steel pipes, hot-rolled steel plates, cold-rolled steel plates, thick plates, and other steel plates and shaped steel, as well as semi-finished products such as slabs that are produced in the process of manufacturing these steel products.

[0032] As shown in Fig. 1, a surface defect detection device 1 according to an embodiment of the present disclosure detects surface defects of a steel material PL as a strip transported on a transport roller table TB in the direction of arrow D in the figure. The surface defect detection device 1 has as its main components lighting devices 2a, 2b, area sensors 3a, 3b, a processor 4, and a display device 5. The lighting devices 2a, 2b correspond to the "lighting section" recited in the claims. The area sensors 3a, 3b correspond to the "imaging section" recited in the claims.

[0033] The lighting devices 2a and 2b have any light-emitting element that illuminates the surface of the strip. For example, the lighting devices 2a and 2b emit light and illuminate the inspection position on the surface of the steel material PL every time the steel material PL moves a certain distance in the direction of arrow D on the transport roller table TB.

[0034] The area sensors 3a and 3b each have an arbitrary light receiving element that detects reflected light from the strip obtained by illuminating the surface of the strip with the lighting devices 2a and 2b. The area sensors 3a and 3b each include, for example, a camera. For example, the area sensors 3a and 3b detect reflected light from an inspection position on the steel material PL surface in synchronization with the light emission of the lighting devices 2a and 2b, respectively, and capture an image of the steel material PL surface.

[0035] 1, the surface defect detection device 1 has two sets of lighting devices and area sensors, but is not limited to this. The surface defect detection device 1 may have one or three or more sets of lighting devices and area sensors depending on, for example, the product width of the steel material PL to be inspected. The surface defect detection device 1 may have multiple lighting devices arranged for one area sensor, or multiple area sensors arranged for one lighting device.

[0036] The arithmetic processing device 4 has one or more processors. In one embodiment, the "processor" is a general-purpose processor or a dedicated processor specialized for a specific process, but is not limited to these. The arithmetic processing device 4 acquires two-dimensional images captured using the area sensors 3a and 3b, and detects surface defects of the steel material PL. The arithmetic processing device 4 has image processing units 41a and 41b and an integrated processing unit 42 built therein.

[0037] The image processors 41a and 41b detect surface defects of the strip by processing a plurality of images including the surface of the strip captured by relatively scanning the surface of the strip using the lighting devices 2a and 2b and the area sensors 3a and 3b. The image processors 41a and 41b execute a series of image processing described later to detect the presence range of the steel material PL reflected in the field of view of each area sensor and surface defects on the steel material PL. The image processors 41a and 41b may determine the type and harmfulness of each defect based on the characteristics of each detected surface defect.

[0038] The integration processing unit 42 aggregates information on the existence range of the steel material PL and information on surface defects detected by the image processing units 41a and 41b, respectively, over the entire steel material PL. The integration processing unit 42 creates inspection performance information such as a defect map showing where defects exist on the steel material PL, and a summary table that tallies the number of defects on the steel material PL by defect type and harmfulness. The integration processing unit 42 outputs the created inspection performance information to the display device 5, and displays it to the operator using the display device 5.

[0039] The display device 5 has one or more output interfaces that output information to present to an operator. For example, the display device 5 has a display that outputs information in the form of a video.

[0040] Next, the image processing executed by the image processing units 41a and 41b will be described with reference to FIG.

[0041] Fig. 2 is a flowchart showing image processing executed by the image processors 41a and 41b in Fig. 1. The image processing procedures executed by the image processors 41a and 41b are the same. Therefore, in the following, image processing in the system of the lighting device 2a, the area sensor 3a, and the image processor 41a will be described as an example.

[0042] In step S1, the image processing unit 41a receives an input image I n Obtain (x,y) as input image I n (x, y) is an image captured by area sensor 3a detecting reflected light from the surface of steel material PL illuminated by lighting device 2a in synchronization with the light emission of lighting device 2a. Here, the subscript n indicates that this is the nth image capture since the start of inspection of steel material PL. (x, y) indicate two-dimensional coordinates in the input image. x is the coordinate corresponding to the width direction of steel material PL, i.e., the direction perpendicular to the direction of arrow D in Figure 1. y is the coordinate corresponding to the traveling direction of steel material PL, i.e., the direction of arrow D in Figure 1. n The (x, y) value represents the luminance at each coordinate (x, y).

[0043] In step S2, the image processing unit 41a converts the input image I n Based on (x,y), the area to be inspected, i.e., the input image I n The range in which the steel material PL exists at (x, y) is detected. The method for detecting the inspection target area may be a method suitable for the inspection target material and its background.

[0044] For example, the input image I nIf the steel material PL in (x,y) is sufficiently bright compared to the background, a threshold T R Using this, a binary image R indicating the inspection target area is generated at each coordinate (x, y) using the following equations (1) and (2). n A method for calculating (x, y) can be used.

[0045] I n (x,y)≧T R When R n (x,y)=1 (area to be inspected) (1) I n (x,y) <T R When R n (x,y)=0 (background) (2)

[0046] In addition, the input image I n Search in the x direction for each y coordinate of (x,y), and n (x,y) is T R It is also possible to employ a method in which the position where a predetermined number of consecutive pixels are exceeded for the first time is recognized as the boundary between the steel material PL and the background, and the inspection target area is detected.

[0047] In step S3, the image processing unit 41a converts the input image I n The luminance of the pixels included in the inspection target area of ​​(x, y) is added to the integrated luminance image S(x, y). That is, the integrated luminance image S(x, y) includes the input images I1(x, y) to I n-1 The luminance values ​​within the inspection target region in each image up to (x, y) are accumulated, and the image processing unit 41a updates the accumulated luminance image S(x, y) as follows: S(x,y):=S(x,y)+I n (x,y)×R n (x,y) (3)

[0048] In step S4, the image processing unit 41a converts the input image I nFor pixels included in the inspection target area of ​​(x, y), add 1 to the count image C(x, y). That is, in the count image C(x, y), the pixel at each coordinate (x, y) is added to the input image I1(x, y) to I n-1 The number of times (x, y) falls within the inspection target region is accumulated, and the image processing unit 41a updates the count image C(x, y) as follows: C(x,y):=C(x,y)+R n (x,y) (4) The image processing unit 41a initializes a count image C(x, y) for each steel material PL that is an inspection target material.

[0049] In step S5, the image processing unit 41a judges whether or not there is an input image that has not been acquired. If the image processing unit 41a judges that there is an input image that has not been acquired, the image processing unit 41a returns to step S1 and acquires the next input image I n+1 The process of steps S1 to S4 is executed again for (x, y). If the image processing unit 41a determines that there is no input image that has not yet been acquired, the process of the next step S6 is executed.

[0050] In step S6, the image processing unit 41a calculates an average image A(x, y) of the multiple input images acquired in step S1. More specifically, the image processing unit 41a calculates the average image A(x, y) by dividing the integrated luminance image S(x, y) by the count image C(x, y). That is, the image processing unit 41a calculates the average image A(x, y) at each coordinate (x, y) as follows: When C(x,y)≠0, A(x,y):=S(x,y) / C(x,y) (5) (However, when C(x,y)=0, the value of A(x,y) is arbitrary.)

[0051] The value of the average image A(x,y) calculated by formula (5) is the average luminance at the position where the steel material PL corresponds to the coordinates (x,y) of the input image within the field of view of the area sensor 3a, and the average image as a whole represents a shading pattern caused by unevenness of the lighting device 2a, etc. Here, in the average image calculation step of step S6, the image processing unit 41a recognizes the inspection target area where the band-like object exists in each of the multiple input images, and allows only pixels in the inspection target area to contribute to the average image A(x,y). In other words, please note that the luminance of the input image contributes to the average value only when the coordinates (x,y) in each input image are within the inspection target area, and contributions from the background are excluded.

[0052] In step S7, the image processing unit 41a performs shading correction on each of the multiple images acquired in step S1 using the average image A(x, y) to acquire a corrected image. More specifically, the image processing unit 41a uses the average image A(x, y) calculated in step S6 to obtain an input image I n Perform shading correction of (x,y).

[0053] In the image correction step of step S7, the image processing unit 41a divides each of the multiple images acquired in step S1 by the average image A(x, y). More specifically, the image processing unit 41a divides the input image I n The image with shading correction of (x,y) is J n If (x, y), the division-type shading correction is calculated as follows: J n (x,y):=I n (x,y) / A(x,y) (6)

[0054] Alternatively, in the image correction step of step S7, the image processing unit 41a subtracts the average image A(x, y) from each of the multiple images acquired in step S1. More specifically, in subtraction-type shading correction, the image processing unit 41a performs calculation according to the following equation. J n (x,y):=I n (x,y)-A(x,y) (7)

[0055] However, the above shading correction calculation is performed on the input image I n Within the inspection area at (x, y), that is, R n It is executed for the coordinates (x, y) where (x, y) = 1, and the background (R n The coordinates corresponding to (x, y) = 0) are set to arbitrary values. With this shading correction, the average luminance level of the healthy part of the steel PL becomes a constant value regardless of the position. For example, in the division type shading correction using formula (6), the average luminance level of the healthy part is J n (x, y) = 1. For example, in the subtractive shading correction using equation (7), the average luminance level of the healthy part is J n (x,y)=0.

[0056] Here, as described above, in the average image A(x, y), the brightness of the input image contributes to the average value only when the coordinates (x, y) in each input image are within the inspection target area, and the contribution from the background is excluded. Therefore, when the steel material PL meanders or changes in width, the series of input images I n Shading correction is performed appropriately even in coordinates (x, y) where the object is either the steel product surface or the background.

[0057] The image processing unit 41a converts the shading-corrected image J n If it is necessary to adjust the average brightness level of (x, y) or the contrast of the defect signal, an appropriate linear conversion or the like may be performed. For example, the image processing unit 41a converts the image that has been subjected to division-type shading correction using equation (6) into a grayscale image with 256 gradations so that the average brightness level becomes 128. In such a case, the image processing unit 41a converts the converted shading-corrected image into a J n '(x,y) and execute the conversion process as follows: J n '(x,y)=J n (x,y)×128 (8)

[0058] In step S8, the image processing unit 41a performs binarization of the shading correction image J obtained in step S7. The binarization is for determining whether each pixel belongs to a defective part or not. If the binarized image is B n (x, y), then for each pixel of the shading correction image J n (x, y), the binarized image B n (x, y) is calculated as follows by comparing with two threshold values T1 and T2 at each pixel of J n (x, y). J n (x, y) ≤ T1 or J n (x, y) ≥ T2, then B n (x, y) := 1(9) T1 < J n (x, y) < T2, then B n (x, y) := 0(10)

[0059] However, the threshold value T1 is a value smaller than the average luminance level. The threshold value T2 is a value larger than the average luminance level. When the shading correction is of the division type, T1 < 1 < T2. When the shading correction is of the subtraction type, T1 < 0 < T2. Also, when the conversion as in equation (8) is performed, the threshold values T1 and T2 are the values before and after the average luminance level after conversion, respectively. Further, if the defect to be detected is either brighter or darker than the healthy part on the image, the image processing unit 41a may perform the comparison with only one of the threshold values T1 and T2.

[0060] In step S9, the image processing unit 41a extracts blobs from the binarized image B n (x, y) obtained in step S8 and labels the extracted blobs. Here, a blob refers to a region where the pixels with the value of 1 in the binarized image B n (x, y), that is, the pixels belonging to the defective part are continuous. Labeling is to assign an identification label (serial number) to each blob.

[0061] The binarized image B nIn (x, y), the pixel regions belonging to the same defect may not be continuous and may be divided into two or more blobs. To deal with such cases, the image processor 41a may regard the blobs as one blob when the distance between the blobs is smaller than a predetermined distance and assign the same label to them. In addition, the image processor 41a may remove blobs with a small area (number of pixels) because small blobs may occur due to harmless patterns on the surface of the steel material PL even in healthy parts.

[0062] In step S10, the image processor 41a calculates the feature amount of the blob labeled in step S9, i.e., the defect. The feature amount is calculated based on the shading-corrected image J n (x,y) (or input image I n (x,y)) and binarized image B n The feature values ​​are calculated for pixel regions belonging to each blob (defect) at (x, y). The feature values ​​include, for example, those expressing the size of the defect, such as width, length, and area, those relating to the shape of the defect, such as aspect ratio and circularity, and those relating to the shading of the defect, such as average brightness and brightness histogram.

[0063] In step S11, the image processing unit 41a outputs the shading-corrected image J n A surface defect of the strip is detected based on (x, y). More specifically, the image processor 41a performs defect determination based on the feature amount of each blob (defect) calculated in step S10. For example, the image processor 41a assigns a type of defect and a degree of harmfulness to each blob. As a method of determination, a predetermined IF-THEN rule may be applied, a discriminator generated by various machine learning methods may be applied, or a combination of these may be applied.

[0064] In step S12, the image processing unit 41a processes the input image I nIt is determined whether (x, y) exists. If the image processing unit 41a determines that there is an input image that has not been completed, the process returns to step S7 and repeats the processes from step S7 to step S11. If the image processing unit 41a determines that the process has been completed for all input images, the process ends.

[0065] By executing the above-mentioned steps S1 to S12, the surface defect detection device 1 can properly perform shading correction near the plate edge even when the steel material PL meanders and the edge position fluctuates, and can realize shading correction that does not cause "folding" around the defective portion. This enables the surface defect detection device 1 to detect surface defects with high sensitivity and appropriately determine the type and harmfulness of the defect.

[0066] In addition, if there is a non-steady part such as a defect part in the above-mentioned inspection target area, that is, a part having a higher or lower brightness than the surrounding healthy parts, it will contribute to the average image A(x,y). For example, the contribution degree becomes large at the coordinate (x,y) where the value of the count image C(x,y) is small, and unnecessary distortion occurs in the average image A(x,y), which is an estimate of the shading pattern. In order to suppress such distortion, the image processing unit 41a performs the following in step S2: n In addition to (x,y), the following binary image Q n You can also calculate (x,y). R n When (x,y)=1 and (x,y) is in the stationary part, Q n (x,y)=1 (11) R n When (x,y)=0 or (x,y) is in a non-stationary part, Q n (x,y)=0(12)

[0067] In step S3, the image processing unit 41a performs calculation by replacing the formula (3) with the following formula (3'). S(x,y):=S(x,y)+I n (x,y)×Q n (x,y) (3')

[0068] In step S4, the image processing unit 41a performs calculation by replacing the formula (4) with the following formula (4'). C(x,y):=C(x,y)+Q n (x,y) (4')

[0069] In order to determine whether a part is a steady part or a non-steady part in the formulas (11) and (12), the image processing unit 41a uses, for example, the input image I n Alternatively, the image processing unit 41a may perform steps S1 to S8 once to extract a shading-corrected image J. n (x,y) and binary image B n After calculating (x,y), B n A pixel (x, y) where (x, y) = 1 may be determined to be a non-steady part. After that, the image processing unit 41a may return to step S1 again, and execute steps S3 and S4 up to step S12 by substituting the above formulas (3') and (4').

[0070] At this time, in the average image calculation step of step S6, the image processing unit 41a may recognize non-steady parts within the inspection area in each of the multiple input images, and contribute only the pixels of the steady parts to the average image A(x, y).

[0071] Hereinafter, examples of the present disclosure and comparative examples based on the conventional technology will be mainly described with reference to FIGS. 3 to 7C.

[0072] FIG. 3 is a diagram showing a plurality of images of the steel material PL in FIG. 1. FIG. 3 shows some of 61 images obtained by repeatedly capturing images of the edge portion E to the center portion of a thick steel plate as the steel material PL being transported on the transport roller table TB at regular intervals in the longitudinal direction of one thick steel plate using a camera of the imaging unit fixed above the transport roller table TB and a flash illumination of the illumination unit. (a) in the upper part of FIG. 3 shows an image near the leading end of the thick steel plate in the longitudinal direction. (b) in the middle part of FIG. 3 shows an image of the longitudinal center portion of the thick steel plate. (c) in the lower part of FIG. 3 shows an image near the longitudinal tail end of the thick steel plate. As shown in these three images, it can be seen that the position of the edge portion E of the thick steel plate moves within the field of view of the camera and meanders on the transport roller table TB.

[0073] Fig. 4 is a diagram showing an example of the present disclosure, in which image processing is applied to a series of images of the thick steel plate in Fig. 3 by the surface defect detection device 1 according to an embodiment of the present disclosure, and shows an example of an input image having a stain-like defect and a processed image thereof.

[0074] (a) of Fig. 4 shows the input image I n FIG. 4(b) is a diagram showing an average image A(x,y). The average image A(x,y) in FIG. 4(b) corresponds to a shading pattern. FIG. 4(c) is a diagram showing a binary image R(x,y) showing the inspection target region. n (d) of FIG. 4 is a diagram showing the shading-corrected image J n (e) of FIG. 4 is a diagram showing the shading-corrected image J (x, y). n B is a binary image obtained by thresholding (x,y). n FIG. 4 is a diagram showing (x, y) of the binary image B in FIG. n (x,y) is the image in which the defect is detected.

[0075] In Fig. 4, the image processing unit 41a searches for the edge of the thick steel plate from the left side of the input image in detecting the inspection target area in step S2 in Fig. 2, and determines the position where the brightness value exceeds a predetermined value for 20 or more consecutive pixels to the right as the edge position. In addition, since halation occurs near the edge, the image processing unit 41a determines that 18 pixels from the edge are non-steady parts and excludes them from the inspection target area. Furthermore, in the labeling process in step S9 in Fig. 2, the image processing unit 41a removes blobs with an area smaller than 20 pixels as noise.

[0076] Fig. 5 is a diagram showing a comparative example according to the conventional technology. Fig. 5 shows the result of image processing in which a shading pattern is calculated using a moving average filter and shading correction is performed on the same input image as the image shown in Fig. 4(a). The shading pattern is calculated for each y coordinate and is performed using a moving average filter of 64 pixels in the x direction.

[0077] FIG. 5(a) is a diagram showing the same input image as FIG. 4(a). FIG. 5(b) is a diagram showing a shading pattern generated by applying a moving average filter to the input image of FIG. 5(a). FIG. 5(c) is a diagram showing the same binary image as FIG. 4(c) showing the inspection target area. FIG. 5(d) is a diagram showing a shading-corrected image based on the shading pattern of FIG. 5(b). FIG. 5(e) is a diagram showing a binary image obtained by subjecting the shading-corrected image of FIG. 5(d) to threshold processing. The binary image of FIG. 5(e) is an image in which a defective portion is detected. For processes other than shading correction, the same processes and parameters as FIG. 4 were used.

[0078] Fig. 6 is a diagram showing a comparative example according to another conventional technique, in which a shading pattern is calculated using a second-order polynomial approximation for an input image identical to the image shown in Fig. 4(a) and then shading correction is performed, showing the result of image processing.

[0079] FIG. 6(a) is a diagram showing the same input image as FIG. 4(a). FIG. 6(b) is a diagram showing a shading pattern generated by applying a quadratic polynomial approximation to the luminance waveform in the x direction for each y coordinate of the input image of FIG. 6(a). FIG. 6(c) is a diagram showing the same binary image as FIG. 4(c) showing the inspection target area. FIG. 6(d) is a diagram showing a shading-corrected image based on the shading pattern of FIG. 6(b). FIG. 6(e) is a diagram showing a binary image obtained by subjecting the shading-corrected image of FIG. 6(d) to threshold processing. The binary image of FIG. 6(e) is an image in which a defective portion is detected. For processes other than shading correction, the same processes and parameters as FIG. 4 were used.

[0080] 7A to 7C compare the results of shading correction between an embodiment based on the technology of the present disclosure and two comparative examples. Fig. 7A is a graph showing a corrected luminance waveform between A and A in the shading-corrected image of Fig. 4(d). Fig. 7B is a graph showing a corrected luminance waveform between E and I in the shading-corrected image of Fig. 5(d). Fig. 7C is a graph showing a corrected luminance waveform between C and C in the shading-corrected image of Fig. 6(d).

[0081] The average image A(x,y) in FIG. 4(b) showing an embodiment of the present disclosure is obtained by multiplying the input image I(x,y) in FIG. n The shading of (x,y) is well represented with almost no effect on the pattern and defects on the steel plate surface. n In (x, y), as shown in the graph of FIG. 7A, uniform brightness is achieved overall, and no "aliasing" occurs around the defect.

[0082] On the other hand, in the shading correction image of Fig. 5(d), which is a comparative example of the conventional technology, there are slightly brighter areas on the left and right of the stain-like defect (dark area), i.e., "aliasing" occurs with respect to the defect (dark area), as shown in the graph of Fig. 7B. Note that, although there is no false detection of the "aliasing" area in this comparative example shown in Fig. 5(d), when the signal of the defect area is stronger, the signal level of the "aliasing" area also becomes proportionally larger, which causes false detection.

[0083] Furthermore, in the shading-corrected image of FIG. 6(d), which is a comparative example of another conventional technique, as shown in the graph of FIG. 7C, although there is no "aliasing" as in FIG. 7B, the right side of the image is somewhat dark. This indicates that the shading correction is not appropriate and has not been performed sufficiently. As a result, a false detection occurs at the right end of the image in the binarized image of FIG. 6(e).

[0084] As described above, by using the technology disclosed herein, compared to conventional technology, shading correction can be performed appropriately even in situations where the material being inspected is meandering or has a change in width, and "folding" around defects is also suppressed.

[0085] According to the surface defect detection method and surface defect detection device 1 of the embodiment described above, an average image calculated from a plurality of images is used as a shading pattern, and only pixels within the inspection target area in each of the plurality of images contribute to the average image, so that it is possible to detect surface defects of a strip-shaped body with higher accuracy. More specifically, "folding" around the defect or "hollowing" of a defect with a large area is suppressed in the corrected image. In addition, even if the edge position shown in the image changes due to meandering or width change of the inspection target material, it is possible to perform shading correction appropriately up to the vicinity of the edge.

[0086] Although the present disclosure has been described based on the drawings and examples, it should be noted that those skilled in the art can make various modifications and alterations based on the present disclosure. Therefore, it should be noted that these modifications and alterations are included in the scope of the present disclosure. For example, the functions included in each configuration or each step can be rearranged so as not to be logically inconsistent, and multiple configurations or steps can be combined or divided into one.

[0087] For example, the shape, size, arrangement, orientation, number, etc. of each of the above-mentioned components are not limited to the above description and the contents illustrated in the drawings. The shape, size, arrangement, orientation, number, etc. of each of the components may be configured arbitrarily as long as the function can be realized.

[0088] For example, a general-purpose electronic device such as a smartphone or a computer can be configured to function as the surface defect detection device 1 according to the embodiment described above. Specifically, a program describing the processing contents for realizing each function of the surface defect detection device 1 according to the embodiment is stored in a memory of the electronic device, and the program is read and executed by a processor of the electronic device. Therefore, the disclosure according to one embodiment can also be realized as a program executable by a processor.

[0089] Alternatively, the disclosure according to an embodiment may be realized as a non-transitory computer-readable medium storing a program executable by one or more processors to cause the surface defect detection device 1 according to the embodiment to execute each function. It should be understood that these are also included within the scope of the present disclosure.

[0090] In the above embodiment and example, the "imaging unit" is configured to use an area sensor, i.e., a camera having imaging elements arranged two-dimensionally, but this is not limiting. The "imaging unit" may also use a line sensor, i.e., a camera having imaging elements arranged one-dimensionally. When a line sensor is used, the inspection target material is imaged line by line in the X direction. Therefore, the surface defect detection device 1 receives an input image I nThe size of the Y direction of (x,y) is set to 1, or the input image I n It is also possible to form (x, y).

[0091] In the above embodiment, the "strip" has been described as including, for example, steel material, but is not limited thereto. The strip may include any strip-shaped or sheet-shaped object other than steel material. [Explanation of symbols]

[0092] 1. Surface defect detection device 2a Lighting device (lighting section) 2b Lighting device (lighting section) 3a Area sensor (imaging section) 3b Area sensor (imaging section) 4 Processing Unit 41a Image processing unit 41b Image processing unit 42 Integrated Processing Unit 5 Display device E Edge part PL steel material (strip) TB Transport Roller Table

Claims

1. 1. A method for optically detecting surface defects on a strip, comprising: an image acquisition step of detecting reflected light from the strip obtained by illuminating the surface of the strip, and capturing an image of the surface of the strip while relatively scanning the surface of the strip to acquire a plurality of images including the surface of the strip; an average image calculation step of calculating an average image of the plurality of acquired images; an image correction step of performing shading correction on each of the acquired images using the average image to obtain a corrected image; a defect detection step of detecting surface defects of the strip based on the corrected image; Including, the step of calculating the average image includes identifying an inspection target area in each of the plurality of images in which the band exists, and allowing only pixels in the inspection target area to contribute to the average image. Surface defect detection methods.

2. 2. The surface defect detection method according to claim 1, The image correction step includes dividing each of the acquired images by the average image. Surface defect detection methods.

3. 2. The surface defect detection method according to claim 1, The image correction step includes subtracting the average image from each of the plurality of acquired images. Surface defect detection methods.

4. 4. The surface defect detection method according to claim 1 , further comprising the steps of: the average image calculation step includes recognizing, in each of the plurality of images, a non-constant part within the inspection target region in the image, and allowing only pixels of the constant part to contribute to the average image. Surface defect detection methods.

5. 4. The surface defect detection method according to claim 1 , further comprising the steps of: The strip includes a steel material. Surface defect detection methods.

6. A surface defect detection device for optically detecting surface defects on a strip, comprising: an illumination unit that illuminates a surface of the strip; an imaging unit that detects reflected light from the strip obtained by illuminating a surface of the strip with the illumination unit; an image processing unit that processes a plurality of images including the surface of the strip captured by relatively scanning the surface of the strip using the illumination unit and the imaging unit to detect surface defects of the strip; Equipped with The image processing unit includes: Calculating an average image of the plurality of images; obtaining a corrected image by performing shading correction on the plurality of images using the average image; detecting surface defects on the strip based on the corrected image; identifying a region of interest in each of the plurality of images in which the band exists, and allowing only pixels in the region of interest to contribute to the average image; Surface defect detection device.

Images