Image correction method and abnormality detection method, and image correction device and abnormality detection device

Abstract

Provided are an image correction method and an abnormality detection method, and an image correction device and an abnormality detection device. The image correction method includes: an acquisition unit that acquires images of surfaces of a plurality of stable metal strip plates, i.e., images captured under different illumination conditions or shooting conditions; a selection unit that selects reference pixels / regions contained in a portion of the metal strip plate from the plurality of images; a determination unit that determines a first image in which a brightness of a certain one of average R values, G values, and B values of the reference pixels / regions is the largest from among the plurality of images; a calculation unit that calculates a respective brightness difference obtained by subtracting average R values, G values, and B values of the reference pixels / regions of all images other than the first image from the average R values, G values, and B values of the reference pixels / regions of the first image; and a correction unit that adds or subtracts an absolute value of the respective brightness difference to / from respective R values, G values, and B values of all pixels of the other all images.

Description

Technical Field

[0001] This invention relates to image correction methods and anomaly detection methods, as well as image correction apparatus and anomaly detection apparatus. Background Technology

[0002] In a strip processing production line, as an example of a technology for reliably detecting attached spacer paper, Patent Document 1 describes a method for detecting spacer paper attached to a strip when it is being transported to the processing production line. The method involves optically detecting the color composition of the strip and detecting the spacer paper based on the ratio between the color components of the detection light and the ratio between the color components of the reference color.

[0003] Existing technical documents

[0004] Patent documents

[0005] Patent Document 1: Japanese Patent Application Publication No. 6-167576 Summary of the Invention

[0006] In rolling production lines, images of the surface of most plates are monitored, and anomalies in the finished plates are determined when a signal exceeding a certain threshold is detected. As an example of such technology, there exists a technique described in Patent Document 1.

[0007] Furthermore, Patent Document 1 discloses the following: When reading the color of a steel plate, variations in the shape and pass order of the steel plate act as interference, causing differences in the ratios of color components even among the same type of steel plate. Therefore, to correct for these differences in the ratios of color components caused by this interference, the color component signals are sampled dozens of times during the movement of the steel plate, and the average value of each color in the sampled color signal is read as a representative value of the reference color.

[0008] Here, brightness varies depending on lighting conditions and is reflected in the image. If brightness changes, even for the same image, the values ​​of R, G, and B will change. Furthermore, depending on the camera's shooting conditions, not only the image brightness but also the values ​​of R, G, and B will vary.

[0009] In this regard, the aforementioned prior art does not consider resetting the representative value of the reference color based on the brightness of the image, thus leaving room for improvement in the accuracy of image processing.

[0010] This invention provides an image correction method and an anomaly detection method, as well as an image correction device and an anomaly detection device, which can improve the accuracy of anomaly detection based on board image processing even under different lighting conditions and / or camera shooting conditions.

[0011] The present application includes a plurality of solutions to the above problems, if one example is cited, characterized in that it comprises the following steps: an acquisition step of acquiring a plurality of images, the image being an image of the surface of a metal strip plate in at least a portion comprising a steady state of rolling by a rolling mill, that is, an image reflecting the metal strip plate taken by a camera at a specified position under different lighting conditions or shooting conditions; a selection step of selecting, for a plurality of images acquired in the acquisition step, reference pixels or reference pixel ranges contained in the portion reflecting the metal strip plate; a determination step of determining, from a plurality of images acquired in the acquisition step, a first image containing a reference pixel or a reference pixel range whose average R value, G value, B value of the reference pixel or the reference pixel range becomes the maximum brightness; a calculation step of calculating the respective brightness differences between the average R value, G value, B value of the reference pixel or the reference pixel range of the first image and the average R value, G value, B value of the reference pixel or the reference pixel range of the same area in all other images except the first image; and a correction step of adding or subtracting the absolute value of each of the brightness differences to each of the R value, G value, B value of all pixel points in all other images, so that the average of each of the R value, G value, B value of the reference pixel or the reference pixel range of all other images is consistent with the average of each of the R value, G value, B value of the reference pixel or the reference pixel range of the first image.

[0012] Inventive Effects

[0013] According to the present application, even in the case where the lighting conditions and / or the shooting conditions of the camera are different, the precision of the anomaly detection based on the image processing of the plate can be improved. The above-mentioned problems, structures and effects will be made clear through the following description of the embodiments. BRIEF DESCRIPTION OF DRAWINGS

[0014] Figure 1 is a diagram showing an overview of a rolling apparatus provided with an image correction device and an anomaly detection device according to an embodiment of the present application.

[0015] Figure 2 is a diagram showing an example of the difference in color tone of images taken of a metal strip plate ((A): first image, (B): second image).

[0016] Figure 3 is a diagram showing the condition of the brightness of the R value, G value or B value of the plate surface in the rolling direction of the image taken of a metal strip plate.

[0017] Figure 4is a graph showing an example of correction of luminance of R value, G value or B value in the rolling direction of the image obtained by photographing the metal strip plate. (A) is a luminance distribution of R value, G value, B value in the rolling direction of the first image and the second image (a case where the luminance difference is positive when the luminance of the first image is subtracted by the luminance of the second image at the reference pixel site). (B) is a luminance distribution of R value, G value, B value in the rolling direction of the first image and the corrected second image.

[0018] Figure 5 is a graph showing an example of correction of luminance of R value, G value or B value in the rolling direction of the image obtained by photographing the metal strip plate. (A) is a luminance distribution of R value, G value, B value in the rolling direction of the first image and the third image (a case where the luminance difference is negative when the luminance of the first image is subtracted by the luminance of the third image at the reference pixel site). (B) is a luminance distribution of R value, G value, B value in the rolling direction of the first image and the corrected third image.

[0019] Figure 6 is a graph showing R value, G value, B value at the point B in Figure 2 .

[0020] Figure 7 is a graph showing R value, G value, B value at the point B in Figure 6 .

[0021] Figure 8 is a graph showing R value, G value, B value at the point B in Figure 2 .

[0022] Figure 9 is a graph showing R value, G value, B value at the point B in Figure 8 .

[0023] Figure 10 is a graph showing an example of a histogram of luminance (R value, G value, B value) of pixels (pixels) in the plate range for one image in the abnormality detection device of the embodiment.

[0024] Figure 11 is a graph of a luminance histogram showing an example of a calculation method of a luminance reference value (R value) in the abnormality detection device of the embodiment.

[0025] Figure 12 is a graph of a luminance histogram and a cumulative value total ratio of a frequency distribution of a calculation method of a luminance reference value (R value) in the abnormality detection device of the embodiment.

[0026] Figure 13 FIG. 2 is a graph of a luminance histogram and a cumulative value total ratio of a frequency distribution thereof, which represents another other example of a calculation method of a luminance reference value (R value) in the abnormality detection device of the embodiment.

[0027] BRIEF DESCRIPTION OF DRAWINGS

[0028] 1: metal strip plate

[0029] 10: F1 stand (rolling mill)

[0030] 11, 21, 31, 41, 51: press-down cylinder

[0031] 12, 22, 32, 42, 52: load detector

[0032] 20: F2 stand (rolling mill)

[0033] 30: F3 stand (rolling mill)

[0034] 40: F4 stand (rolling mill)

[0035] 50: F5 stand (rolling mill)

[0036] 55: rack

[0037] 61, 62, 63, 64, 65: camera

[0038] 71, 72, 73, 74: loop

[0039] 80: abnormality detection device

[0040] 81: image correction section

[0041] 81a: acquisition section

[0042] 81b: selection section

[0043] 81c: decision section

[0044] 81d: calculation section

[0045] 81e: correction section

[0046] 82: threshold operation section

[0047] 83: abnormality detection section

[0048] 85: control device

[0049] 87: monitor

[0050] 90: communication line

[0051] 100: rolling apparatus DETAILED DESCRIPTION

[0052] Using Figures 1 to 12 Embodiments of an image correction method and an anomaly detection method of the present application, and an image correction device and an anomaly detection device will be described. In addition, in the drawings used in this specification, the same or similar reference signs are assigned to the same or corresponding structural elements, and repeated description of these structural elements is omitted.

[0053] First, using Figure 1 The overall structure of a rolling apparatus including the image correction device and the anomaly detection device will be described. Figure 1 is a schematic diagram showing the structure of the image correction device and the anomaly detection device of the present embodiment and a rolling apparatus provided with these devices.

[0054] Figure 1 The rolling apparatus 100 shown in FIG. 1 is a rolling apparatus that rolls a metal strip 1, and is provided with an F1 stand 10, an F2 stand 20, an F3 stand 30, an F4 stand 40, an F5 stand 50, cameras 61, 62, 63, 64, 65, loops 71, 72, 73, 74 for tension control, an anomaly detection device 80, a control device 85, a monitor 87, and the like. In addition, the F1 stand 10, the F2 stand 20, the F3 stand 30, the F4 stand 40, the F5 stand 50, the cameras 61, 62, 63, 64, 65, the anomaly detection device 80, and the control device 85 are connected by a communication line 90.

[0055] In addition, regarding the rolling apparatus 100, the configuration shown in FIG. 1 is not limiting, and the number of stands can be two or more. Figure 1 The configuration shown in FIG. 1 in which five stands are provided is the minimum, and two or more stands are sufficient. In addition, the apparatus to which the image correction method and the anomaly detection method of the present application, and the image correction device and the anomaly detection device are applied need not be a rolling apparatus having two or more stands, and can be a single stand rolling mill.

[0056] The F1 stand 10, the F2 stand 20, the F3 stand 30, the F4 stand 40, and the F5 stand 50 each have an upper work roll and a lower work roll, an upper backup roll and a lower backup roll that are supported by being brought into contact with the upper work roll and the lower work roll, respectively, a press-down cylinder 11, 21, 31, 41, 51 provided at the upper portion of the upper backup roll, and a load detector 12, 22, 32, 42, 52. In addition, a six-stage structure in which an intermediate roll is further provided between each work roll and each backup roll can be provided.

[0057] The loop 71 is a tension control roll provided between the F1 stand 10 and the F2 stand 20. The loop 71 is disposed so as to extend in the width direction of the metal strip 1 with its rotation axis, and is provided so as to hold the metal strip 1 by lifting it upward as it travels.

[0058] Further, the loop 71 can be considered as a device that is lifted upward by a force applied upward, for example, by a spring or the like, by a hydraulic cylinder or a motor, or the like.

[0059] The camera 61 is disposed to take an image of the plate surface of the rolled metal strip plate 1 at the position periphery between the F1 stand 10 and the F2 stand 20 (the position periphery where the loop 71 for tension control is provided). The data of the image taken by the camera 61 is transmitted to the abnormality detection device 80 via the communication line 90.

[0060] Similarly, the loop 72 for tension control is provided between the F2 stand 20 and the F3 stand 30, the loop 73 for tension control is provided between the F3 stand 30 and the F4 stand 40, and the loop 74 for tension control is provided between the F4 stand 40 and the F5 stand 50.

[0061] Further, the camera 62 is disposed at a position to take an image of the plate surface of the metal strip plate 1 at the lifted periphery region of the metal strip plate 1 lifted upward by the loop 72, the camera 63 is disposed at a position to take an image of the plate surface of the metal strip plate 1 at the lifted periphery region of the metal strip plate 1 lifted upward by the loop 73, the camera 64 is disposed at a position to take an image of the plate surface of the metal strip plate 1 at the lifted periphery region of the metal strip plate 1 lifted upward by the loop 74, and the camera 65 is disposed at a position to take an image of the plate surface of the metal strip plate 1 at a position downstream of the F5 stand 50. The data of the images taken by the cameras 62, 63, 64, 65 is transmitted to the abnormality detection device 80 via the communication line 90.

[0062] The disposed positions of the cameras 61, 62, 63, 64, 65 are preferably outside in the plate width direction of the metal strip plate 1 when viewed from above the metal strip plate 1, and the taking positions of the above cameras are downstream of the F1 stand 10, the F2 stand 20, the F3 stand 30, the F4 stand 40, and the F5 stand 50. In the case where there is a rolling mill downstream of the rolling mill, it is desirable to be at a position intermediate between two rolling stands (at a position approximately the same as the position of the loop), and in the case where there is no rolling mill downstream of the rolling mill, it is desirable to be at a position immediately downstream of the last rolling stand.

[0063] Illumination can be further provided to illuminate the lifted taking region of the metal strip plate 1 lifted upward by the roll, which is mainly taken by the cameras 61, 62, 63, 64, and the taking region of the metal strip plate 1 downstream of the F5 stand 50 (in the case where there is no loop downstream of the rolling mill). The illumination can be ordinary illumination appropriately provided to the ceiling or the like of the rolling plant where the rolling apparatus 100 is provided, and no special new type of illumination apparatus is required in the present application, but dedicated illumination such as a rod-shaped light source can also be provided.

[0064] The abnormality detection device 80 is a device that performs various processes for determining whether or not the plate shape of the metal strip plate 1 is abnormal based on images captured by the cameras 61, 62, 63, 64, 65, and has an image correction section 81, a threshold operation section 82, and an abnormality detection section 83.

[0065] The image correction section 81 is a section that acquires a plurality of images, which are images of the surface of the metal strip plate 1 in a steady state after rolling by the rolling apparatus 100, that is, images captured under different illumination conditions or imaging conditions, and is suitable as a main body of execution of an acquisition process.

[0066] Here, it can be configured so that the plurality of images acquired by the acquisition section 81a are adjusted so that the respective differences in luminance between the average luminance (hereinafter referred to as R value, G value, B value) of the reference pixel points or the reference pixel range of the first image described later and the R value, G value, B value of the average of the above-described reference pixel points or the above-described reference pixel range of the first image other than the first image become 51 or less.

[0067] Further, the "surface of the metal strip plate in a steady state" referred to here is the surface of the plate of the metal strip plate 1 in a state in which normal rolling is performed at a flat portion of the plate that is not bent in the rolling direction.

[0068] In addition, "different illumination conditions or imaging conditions" means a case in which one or more of the shape, luminance, illuminance, color tone, and setting position of the illumination provided in the facility in which the rolling apparatus 100 is provided, or one or more of the setting position and camera angle, aperture value, shutter speed, ISO sensitivity, and digital setting of the imaging element of the cameras 61, 62, 63, 64, 65 are different.

[0069] As to "acquiring a plurality of", either of a manner in which images that have been captured in advance to represent the metal strip plate 1 is acquired from a storage device, or a manner in which image data captured by the cameras 61, 62, 63, 64, 65 is input from the cameras 61, 62, 63, 64, 65 can be adopted, and there is no particular limitation.

[0070] The selection section 81b is a section that selects a portion that contains reference pixel points or a reference pixel range that shows a portion of the metal strip plate 1 from one of the plurality of images acquired by the acquisition section 81a, and is suitable as a main body of execution of a selection process. The number of points of the pixel points that constitute the selected reference pixel range is not particularly limited, and can be several points to several hundred points or so.

[0071] Here, the selection section 81b can select the reference pixel point or the reference pixel range from a portion showing the metal strip plate 1 other than the reflected light region 1A. In addition, it is desirable to set the selected reference pixel point or the reference pixel range as a region where the difference between the brightness of the prescribed range around it and the brightness thereof can be considered to be less than a prescribed value, for example, as described later. Figure 3 As shown, it is desirable to select a region that can be considered to be away from the reflected light region from among the portion showing the brightness distribution in the rolling direction of the plate surface of the metal strip plate 1.

[0072] To this end, the selection section 81b, for example, performs image processing on an image including the lifted region of the metal strip plate 1, and determines a range including the boundary on the upstream side / downstream side with respect to the rolling direction of a portion where the brightness of the reflected light of the plate surface appearing in the image is greater than a certain brightness value as the reflected light region.

[0073] Here, the reflected light region is a region where the reflected light is stronger than in other portions in the range where the metal strip plate 1 appears, and is mainly exhibited by the lifted portions of the loops 71, 72, 73, 74. The cameras 61, 62, 63, 64, 65 are arranged so as to include the reflected light region in the captured image.

[0074] For example, a region where the maximum value of the brightness of the R value, G value, B value of the image of the metal strip plate 1 is 0.9 times or more of the maximum value can be determined as the reflected light region. In addition, it can be set so that when the maximum value of the brightness of the R value, G value, B value is 255, a region where the brightness is 230 or more is considered to be the reflected light region, and when the maximum value of the brightness is less than 255, a region where the brightness is 0.9 times or more of the maximum value thereof is considered to be the reflected light region. However, the definition of the reflected light region is not limited to the above definition, and can be appropriately changed.

[0075] The determination section 81c is a portion that determines a first image from among the plurality of images acquired by the acquisition section 81a, which is suitable to become the execution subject of the determination process, where the first image is an image where the brightness of one of the R value, G value, B value of the average of the reference pixel point or the reference pixel range is the greatest. In the determination section 81c, as the average of the reference pixel point or the reference pixel range, it can be set to include several points to several hundred points or so of pixel points centered on a certain pixel point, and is not particularly limited.

[0076] The calculation section 81d is a portion that calculates each brightness difference between the R value, G value, B value of the average of the reference pixel point or the reference pixel range of the first image and the R value, G value, B value of the average of the reference pixel point or the reference pixel range of all other images other than the first image, which is suitable to become the execution subject of the calculation process.

[0077] The correction section 81e is a portion that adds or subtracts the absolute value of each luminance difference calculated by the calculation section 81d to or from each R value, G value, and B value in all the pixels of the other entire image, so that the portion in which each R value, G value, and B value of the average of the reference pixel or the reference pixel range of the other entire image coincides with each R value, G value, and B value of the average of the reference pixel or the reference pixel range of the first image becomes the main body of execution of the correction process.

[0078] In the correction section 81e, in the R value, G value, and B value after the correction of the luminance difference, when the luminance exceeds 255, the luminance of the coinciding R value, G value, and B value can be corrected to 255.

[0079] The threshold operation section 82 is a portion that calculates various threshold values using the R value, G value, and B value of the first image with respect to the image corrected by the image correction section 81, and becomes the main body of execution of the threshold operation process. In addition, in the case of repeating the rolling, if the same first image is continuously used, it is not necessary to newly calculate various threshold values during that time, and the same value can be continuously used.

[0080] The abnormality detection section 83 is a portion that detects the abnormality of the surface of the metal strip plate 1 based on the threshold values set by the threshold operation section 82 and the image photographed by the cameras 61, 62, 63, 64, 65 and corrected in luminance by the image correction section 81, and becomes the main body of execution of the abnormality detection process.

[0081] The abnormality detection section 83 in the present embodiment, for example, with respect to the reflected light region within the image, divides the luminance data of the pixels within the range of the metal strip plate 1 into R values, G values, and B values, subtracts the reference R value from the above R value, subtracts the reference G value from the above G value, subtracts the reference B value from the above B value, calculates each luminance difference, and detects as a rolling abnormality in the case where both of the above luminance differences of the two components in each of the above luminance differences are each above each luminance difference threshold value.

[0082] The abnormality detection process by the abnormality detection section 83 is desirably executed at the time of rolling of the metal strip plate 1 by the rolling apparatus 100.

[0083] In addition, the correction process of the image by the image correction section 81 is desirably executed after the maintenance of the rolling apparatus 100 or after the maintenance of the lighting apparatus, and in particular, after the change of the photographing conditions such as the setting position or the camera angle of the cameras 61, 62, 63, 64, 65, the exposure, the exposure time, the white balance, the digital setting of the camera element, and the like.

[0084] On the contrary, the process of calculating the threshold value by the threshold operation section 82 is desirably executed after the change of the first image, and in the case of continuously using the same first image, it is not necessary to particularly execute since the threshold value calculated in advance is used at the time of rolling of the metal strip plate 1 by the rolling apparatus 100.

[0085] The control device 85 is a device that controls the operation of each device in the rolling apparatus 100, and in the present embodiment, is a device that performs various controls corresponding to the judgment of the plate shape of the metal strip plate 1 in the abnormality detection section 83 of the abnormality detection device 80.

[0086] These abnormality detection device 80 and control device 85 can be configured by a computer having a monitor 87 such as a liquid crystal display, an input device, a storage device, a CPU, a memory, and the like, can be configured as a device by one computer, can be configured by different computers, and are not particularly limited.

[0087] The control of the operation of each device by the abnormality detection device 80 and the control device 85 is performed based on various programs recorded in the storage device. Furthermore, the control processing of the operation performed by the abnormality detection device 80 and the control device 85 can be summarized in one program, can be each dispersed in a plurality of programs, or can be a combination thereof. In addition, a part or all of the program can be implemented in a dedicated hardware, or can be modularized.

[0088] The monitor 87 is a display device such as a display or an acoustic device such as an alarm, and is a device for communicating to an operator about a disposal work when the abnormality detection device 80 judges that a problem such as a plate shape defect or plate crushing has occurred, and therefore a display is mostly used as such a monitor 87.

[0089] Here, the above-described abnormality detection device 80 includes a display signal section that transmits a signal about the content displayed on the monitor 87 to the monitor 87.

[0090] The operator can confirm the state of the plate shape by visually observing the display screen of the monitor 87 and each stand itself and between each stand during the work.

[0091] Furthermore, it is not limited to the mode in which the control device 85 automatically performs the operation of communicating to the operator the occurrence of the plate shape defect and improving the plate shape defect, and can be configured in a mode in which only the display on the monitor 87 is performed, a mode in which the display to the monitor 87 is omitted and only the operation of improving the plate shape defect is automatically performed by the control device 85. Furthermore, it is the same in the case where the plate crushing occurs, not only the occurrence of the plate shape defect.

[0092] Next, a specific example of the flow of the correction processing of the image in the image correction section 81 in the present application (the image correction method) will be described using Figure 2 and the subsequent drawings.

[0093] Figure 2 is a schematic view showing an example of a case where the image obtained by photographing the metal strip plate 1 has a different color tone.

[0094] Figure 2 In the metal strip 1 shown, Figure 2 Image (A) shows a schematic diagram of the image (first image) obtained by actually photographing the metal strip 1. Figure 2 Image (B) shows a schematic diagram of images (image 2) obtained by taking the original image with a camera under different shooting conditions.

[0095] like Figure 2 As shown in the diagram, in the first and second images, although both are of the same subject, the brightness and hue of the images are slightly different.

[0096] The following shows the use of and Figure 2 The results are obtained by checking the brightness of the pixels that make up the image using the same actual image as the schematic diagram shown. In the first and second images, A1 and A2 (point A: the position outside the range of metal strip 1 = background), B1 and B2 (point B: the stable surface position within the range of metal strip 1 that is not in the reflected light area 1A), C1 and C2 (point C: the position within the reflected light area 1A within the range of metal strip 1), and D1 and D2 (point D: the stable surface position within the range of metal strip 1 that is not in the reflected light area 1A) are the same parts of the image, but the hue is not the same in the first and second images. Therefore, for each point, the actual brightness (R value, G value, B value) of the two images is checked.

[0097] The brightness (R value, G value, B value) of each point A (A1, A2), B (B1, B2), C (C1, C2), and D (D1, D2) in the first and second images, as well as the differences in brightness between the first and second images at points A, B, C, and D (A1-A2, B1-B2, C1-C2, D1-D2) and their average values ​​are shown in Table 1 below.

[0098] Table 1

[0099] A1 A2 A1-A2 B1 B2 B1-B2 C1 C2 C1-C2 D1 D2 D1-D2 R value 68 11 57 190 134 56 255 237 18 185 119 66 G value 65 16 49 145 102 43 254 249 5 142 97 45 B value 58 19 39 116 77 39 250 247 3 125 74 51 Average 64 15 49 150 104 46 253 244 9 151 97 54

[0100] From the above measurements of R, G, and B values, it can be seen that, except for the positions within the reflected light area (points C: C1, C2), the R, G, and B values ​​within the range of the metal strip 1 and the R, G, and B values ​​of the background (outside the range of the metal strip 1), i.e., the brightness differences between the first image and the second image at points A, B, and D, are close to each other.

[0101] Therefore, the R value, the G value, and the B value of all the pixel points of the configuration image, which become the basis of the image processing for the abnormality detection, are corrected in advance by the following method in which the hue of the steady surface position (for example, the B2 point or the D2 point) in the range of the metal strip 1 of the second image is made to coincide with the hue of the steady surface position (for example, the B1 point or the D1 point) in the range of the metal strip 1 of the first image, and then the abnormality detection processing based on the image analysis is performed. By this processing method, the image analysis can be performed using the same threshold value at all times.

[0102] In addition, if the brightness variation of the plate surface in the rolling direction of the metal strip 1 appearing in the image is schematically shown, as shown in FIG. 6. Figure 3 Figure 3 is a graph showing the condition of the brightness variation of the R value, the G value, or the B value in the rolling direction of the image taken at the loop portion between the rolling stands. At the loop portion, since the metal strip 1 is bent in the rolling direction (toward the height direction), there is a reflected light region showing a higher brightness than other portions. In addition, at the position of the steady surface of the metal strip 1 in which the metal strip 1 is not bent in the rolling direction, the B point or the D point in the reference pixel point or the reference pixel range is set. Figure 2 In the image taken at the position other than the loop portion between the rolling stands, that is, for example, downstream of the last rolling stand, the position of the steady surface of the metal strip is set as the reference pixel point or the reference pixel range immediately downstream of the last rolling stand.

[0103] Here, since the maximum value of the brightness of the R value, the G value, and the B value is 255, the brightness is high in the reflected light region (the C point and the region corresponding to the vicinity thereof) regardless of how the image is taken, and there is a case where the value of the brightness reaches the limit of 255. That is, the maximum value at the reflected light region is R = 255, G = 255, and B = 255 (white).

[0104] For example, in the case of the brightness data shown in Table 1, the above correction processing is performed as follows.

[0105] First, the hue of the position (the B point: B1, B2) in the range of the metal strip 1 is always corrected and fixed to the RGB value (R = 190, G = 145, B = 116) of the first image (B1) in any image.

[0106] That is, the image is processed so that the hue at the position B2 point in the range of the metal strip 1 of the second image coincides with the hue of the B1 point of the first image. In this case, the B point becomes the position of the reference pixel point or the reference pixel range.

[0107] ​Next, the luminance difference of the R value, the G value, and the B value at the position (B point: Bl, B2) in the range of the metal strip plate 1 of the two images (the 1st image and the 2nd image) is checked. In a case where one or all of the R value, the G value, and the B value of the luminance of the 1st image is / are larger than one or all of the R value, the G value, and the B value of the luminance of the 2nd image, the absolute value of the luminance difference of the R value or the G value or the B value that is judged to be large is added to the corresponding R value or G value or B value of the pixel points of the 2nd image as a whole. In a case where one or all of the R value, the G value, and the B value of the 1st image is / are smaller than one or all of the R value, the G value, and the B value of the 2nd image, the absolute value of the luminance difference of the R value or the G value or the B value that is judged to be small is subtracted from the corresponding R value or G value or B value of the pixel points of the 2nd image as a whole.

[0108] Figure 4 and Figure 5 is a diagram schematically showing an example of the correction processing of the luminance of the R value, the G value, or the B value in the rolling direction of the image taken of the metal strip plate 1.

[0109] For example, as shown in Figure 4 , the difference of each luminance of the average R value, the average G value, and the average B value of the reference pixel points or the reference pixel range in the 1st image and the 2nd image is a positive value, and it is considered that the value exceeding the maximum value (255) in the 2nd image after the addition of the absolute value of the above luminance difference. In this case, it is desirable to correct the value exceeding 255 in the 2nd image after the correction to 255.

[0110] In addition, for example, as shown in Figure 5 , the difference of each luminance of the average R value, the average G value, and the average B value of the reference pixel points or the reference pixel range in the 1st image and the 3rd image is a negative value, and in the 3rd image after the subtraction of the absolute value of the above luminance difference, each R value, G value, and B value showing white in the reflected light region is reduced from 255, and there can be a case where the maximum value of the luminance in the reflected light region is significantly lower than 255, as shown in the 3rd image after the correction. In this case, it is considered that there is a possibility that the reflected light region cannot be correctly judged.

[0111] As Figure 5 , there is a possibility that the reflected light region cannot be correctly judged in the correction processing of the luminance of the reflected light region by the subtraction operation of the R value, the G value, and the B value of the luminance of the image. Therefore, in order to avoid the difference of each luminance of the average R value, the average G value, and the average B value of the reference pixel points or the reference pixel range from becoming negative as much as possible, it is desirable to determine the image of the range of the metal strip plate 1 in which the luminance of the stable surface is relatively high as the 1st image.

[0112] That is, in the case where the metal strip plate 1 is a hot-rolled steel plate, it is preferable to adopt as the first image an image in which the brightness of the R value in the color tone of the metal strip plate 1 among the brightness of the reference pixel or the average brightness of the reference pixel range shows the maximum value from past results. The use Figures 6 to 8 The R value of the maximum results is explained in detail.

[0113] Figure 6 Ten cases are randomly selected from image data of the actual rolling of the metal strip plate 1 in the actual rolling apparatus 100, and the R value, the G value, and the B value of the region corresponding to the above-mentioned B point are checked. In addition, the image data of the above-mentioned No. 8 image is adopted as the first image, and the brightness difference of the R value, the G value, and the B value at the B point of the other images is shown. Figure 7 The image data of the above-mentioned No. 8 image is adopted as the first image, and the brightness difference of the R value, the G value, and the B value at the B point of the other images is shown.

[0114] The results of the image data shown in Fig. 8 are that the R value of the No. 8 image shows the maximum value. Figure 6 For example, if the No. 8 image shown in Fig. 8 is set as the first image, the brightness at the above-mentioned B point is R value = 210. In this case, since it is expected that the brightness difference with the other images will not be too large, it is preferable that the brightness at the above-mentioned B point of the other multiple images of the metal strip plate 1 acquired by the acquisition section 81a is selected from images entering the range of R value = 159 to 210.

[0115] Figure 6 That is, from the results of the above-mentioned No. 8 image, it is considered that it is preferable that the brightness difference of the R value between the R value of the brightness of the reference pixel point of the first image or the average R value of the brightness of the reference pixel range and the R value of the other multiple images is 51 or less. In the case where the range (R value = 159 to 210) is deviated, it is preferable that the brightness of the image is adjusted in advance so that the R value enters the range of 159 to 210 in the setting at the time of the imaging of the cameras 61, 62, 63, 64, and 65.

[0116] That is, from the results of the above-mentioned No. 8 image, it is considered that it is preferable that the brightness difference of the R value between the R value of the brightness of the reference pixel point of the first image or the average R value of the brightness of the reference pixel range and the R value of the other multiple images is 51 or less. In the case where the range (R value = 159 to 210) is deviated, it is preferable that the brightness of the image is adjusted in advance so that the R value enters the range of 159 to 210 in the setting at the time of the imaging of the cameras 61, 62, 63, 64, and 65. Figure 9 In addition, in the case where the G value or the B value is larger than the R value, the component having the highest brightness value is set as the selection reference of the first image, but in the case where the rolling apparatus 100 is for hot rolling, since the R value is basically high in the color tone of the metal strip plate 1, it is expected that the R value is set as the selection reference of the first image.

[0117]

[0118] ​​If this is the result, the R, G, and B values ​​of point B in image No. 8 of the first image will be used as a reference. The R, G, and B values ​​of points B in other images will be calculated in such a way that they are the same as those of point B in image No. 8. The brightness difference between the R, G, and B values ​​at point B and the brightness difference at point B will be calculated. The absolute value of the brightness difference will be added to or subtracted from the brightness of all pixels in other images. This will correct the hue of the metal strip 1 at point B in other images so that it is always the same as the hue of the metal strip 1 at point B in image No. 8.

[0119] This processing method ensures that the threshold (parameter) for image processing always uses the same value, without needing to change the threshold based on the image.

[0120] For point D above, try to check whether the above processing method is correct. Figure 8 This image was obtained by checking the R, G, and B values ​​of point D using the same image data as described above. Additionally, Figure 9 This is a graph that sets the aforementioned image No.8 as the first image and shows the brightness difference of the R, G, and B values ​​at point D compared to other images.

[0121] Such as Figure 8 As shown, at point D above, image No. 8 has the highest R value, and its value is also roughly the same as the R value at point B above.

[0122] Moreover, such as Figure 7 and Figure 9 As shown, among the previously randomly selected 10 images, the difference in brightness between the R, G, and B values ​​of each image and that of image No. 8 was examined at points B and D. The results show that the difference in brightness at point D is significant. Figure 9 The maximum value is 51. In the case of image No. 8, the R, G, and B values ​​of the hue of metal strip 1 are approximately the same at points B and D. Therefore, it can be seen that the reference pixel or reference pixel range can be selected from the position of the surface of metal strip 1 outside the reflected light area, i.e., the stable state of the metal strip 1. In addition, according to Figure 9 Based on practical experience, if the brightness difference between the R values ​​of the brightness of the reference pixels of the first image and the other multiple images is 51 or less, or the average R value of the brightness of the reference pixel range, then there is no problem from the practical experience, and it is preferable to be 51 or less.

[0123] Therefore, it can be said that if the position of the surface of the metal strip plate 1 in the steady state of the metal strip plate 1 selected as the first image is set as a reference pixel point or a reference pixel range, the R value, the G value, and the B value in the position or the range are used as a reference, and the luminance of the pixel points of the entire image is corrected in such a manner that the luminance in the reference pixel point or the reference pixel range is always fixed to the same R value, G value, and B value for other images, the value of the threshold value (parameter) set at the time of image processing can always be a fixed value without changing the threshold value setting.

[0124] Next, the drawing after the above description will explain an example of a method of calculating the threshold value in the threshold value operation section 82. Figure 10 Figure 10 is a drawing showing an example of a histogram of the luminance (R value, G value, and B value) of the pixel points in the plate range for one image, Figures 11 to 13 is a drawing showing an example of a method of calculating the luminance reference value associated with the threshold value (parameter) set at the time of image processing of the plate determined to be normal.

[0125] First, it is preferable that the image containing the metal strip plate 1 is acquired by the cameras 61, 62, 63, 64, and 65 at the time of installation of the rolling device 100 or after the maintenance of the rolling device 100, particularly after the change of the shooting conditions of the cameras 61, 62, 63, 64, and 65.

[0126] In the abnormality detection device 80, in the threshold value operation section 82, the range of the metal strip plate 1 is extracted from the luminance region equal to or higher than a certain threshold value in such a manner that the image showing the metal strip plate 1 at the time of rolling after the correction processing by the image correction section 81 is separated into the background and the presence region of the metal strip plate 1 by the binarization processing.

[0127] Next, in the threshold value operation section 82, the luminance distribution of the pixel points inside the range of the metal strip plate 1 is calculated, and the luminance reference value used for the determination of the plate in which no abnormality such as plate crushing occurs is calculated. Here, it is set that the threshold value operation section 82 divides the luminance data into three components of the R value, the G value, and the B value, and calculates three values of the reference R value (R0 value), the reference G value (G0 value), and the reference B value (B0 value) as the luminance reference value.

[0128] Further, it can be set that two values (one of the combinations of the R0 value and the G0 value, the R0 value and the B0 value, and the B0 value and the G0 value) among the above luminance reference values are calculated.

[0129] ​Specifically, the threshold calculation unit 82 calculates, for an image captured by cameras 61, 62, 63, 64, and 65, the luminance distribution (histogram) of each R, G, and B value among all extracted pixels, expressed as the number of measurement points for R, G, and B values. Here, the R, G, and B values ​​are all taken as values ​​in the range of 0 to 255. Figure 10 The image shows an example of a histogram of brightness (R value, G value, B value) for a plate range of an image.

[0130] Next, in the threshold calculation unit 82, as follows Figure 11 As shown, taking the R value of an image as an example, a histogram for each brightness of all pixels is obtained. Within the range of X% brightness above and Y% brightness below the low digits of the histogram, the R value can be determined as the brightness reference value for judging a panel without anomalies. Furthermore, the R value of the mode (Z% brightness from the low digits of the histogram) with the highest number of pixels of the same brightness in the aforementioned histogram can be determined as the brightness reference value for judging a panel without anomalies. For example, as... Figure 12 As shown, the R-value, which represents the area divided into two equal parts by the integral value of the frequency distribution of the R-value (the total ratio of the integral values ​​of the frequency distribution of luminance is 50%), is set as the luminance reference value for judging a board without abnormalities. Alternatively, a value can be selected from an allowable range of ±20% relative to the total ratio of the integral values ​​of the frequency distribution of luminance, and set as the luminance reference value for the R-value. Furthermore, the allowable range does not need to be "±20%" and can be appropriately changed. This method can also be applied to G and B values.

[0131] The value used to determine the brightness reference value (the sum ratio of the frequency distribution integral values ​​of brightness) can be specified by the operator before rolling the metal strip 1 of the abnormal detection object, or it can be a preset value of the device, or it can be a value that is appropriately learned by machine learning.

[0132] Furthermore, it is possible to replace the value of area bisection shown by the integral value of the frequency distribution of the aforementioned R value, and as... Figure 13 As shown, the brightness of the panel with the most measurement points in the frequency distribution of the R value (the mode, i.e., the total percentage of the integral values ​​of the brightness frequency distribution, Z%) is set as the brightness reference value for judging a panel without abnormalities. Furthermore, a value within a tolerance range of ±30% relative to the aforementioned mode can be selected as the brightness reference value for the R value. However, the tolerance range does not need to be "±30%" and can be appropriately changed. This method can also be applied to G and B values.

[0133] Next, the effects of this embodiment will be explained.

[0134] The image correction section 81 of the present embodiment described above has: an acquisition section 81a that acquires a plurality of images that are images of the surface of the steady-state metal strip 1 after rolling by the rolling apparatus 100, that is, images taken under different lighting conditions or imaging conditions; a selection section 81b that selects a reference pixel point or a reference pixel range included in a portion of the metal strip 1 from one of the plurality of images acquired by the acquisition section 81a; a determination section 81c that determines a first image in which the brightness of one of the average R value, G value, and B value of the reference pixel point or the reference pixel range is the largest from among the plurality of images acquired by the acquisition section 81a; a calculation section 81d that calculates each of the brightness differences between the average R value, G value, and B value of the reference pixel point or the reference pixel range of the first image and the average R value, G value, and B value of the reference pixel point or the reference pixel range of all other images except the first image; and a correction section 81e that adds or subtracts the absolute value of each of the brightness differences to each of the R value, G value, and B value of all the pixel points of all other images so that each of the average R value, G value, and B value of the reference pixel point or the reference pixel range of all other images matches each of the average R value, G value, and B value of the reference pixel point or the reference pixel range of the first image.

[0135] Thus, all of the plurality of images can be simulated as images taken under the same lighting conditions and imaging conditions of the camera. Therefore, even for images taken under different lighting conditions and / or imaging conditions, the same various thresholds determined for judgment can be used in image processing, so that it is not necessary to newly determine the various thresholds for abnormality judgment each time the lighting conditions and imaging conditions are changed, and image processing can be performed quickly and stably.

[0136] In addition, especially in the case of the image of the metal strip 1 at the loop portion between the rolling mill stands, a reflected light region in which the R value, G value, and B value are close to white close to the maximum value (255) is reflected, and the color tone of the metal strip 1 is hardly properly shown. Therefore, by selecting the reference pixel point or the reference pixel range from a position on the surface of the steady-state metal strip 1 other than the portion of the metal strip 1, that is, the reflected light region, the selection section 81b can correct the color tone of the reference pixel point or the reference pixel range of all other images to the same color tone. Furthermore, in the case where there is no rolling mill on the downstream side of the rolling mill, by selecting the reference pixel point or the reference pixel range from a position immediately on the downstream side of the last rolling mill stand, the color tone of the reference pixel point or the reference pixel range of all other images can be corrected to the same color tone.

[0137] Further, in a case where the luminance difference of the R value in the reference pixel or the reference pixel range exceeds 51, in the correction process, the value of the reflected light region is affected by a case where the R value, the G value, and the B value all exceed 255, and thus the area of the reflected light region that shows white is likely to become large. Therefore, by providing the acquisition unit 81a to adjust the acquired plurality of images so that the luminance difference becomes 51 or less, it is possible to suppress a decrease in the accuracy of the determination of the reflected light region.

[0138] Further, the correction unit 81e corrects the luminance value of the R value, the G value, and the B value to 255 when the luminance value of the R value, the G value, and the B value after the addition or subtraction of the absolute value of the luminance difference exceeds 255, and thus it is possible to set the luminance value of the pixel to the normal value range corresponding to the case where the determinable value range is 0 to 255, and it is possible to perform more accurate correction processing.

[0139] Further, the abnormality detection device 80 can calculate the threshold value of the R value, the G value, and the B value that indicates the determinability for the image obtained using the illumination condition and the imaging condition used in the first image, apply the threshold value used in the first image to the corrected image that takes the illumination condition and the imaging condition into consideration, and detect the abnormality of the surface of the metal strip 1, and thus it is possible to improve the accuracy of the detection of the abnormality of the metal strip 1 without calculating various threshold values again. Further, since it is not necessary to calculate the threshold value multiple times, it is possible to reduce the burden on the operator during the operation.

[0140] <Other>

[0141] Further, the present application is not limited to the above-described embodiments, and various modifications and applications are possible. The above-described embodiments are described in detail in order to easily understand the present application, but are not limited to necessarily include all the structures described.

Claims

1. A method of modifying an image, characterized by, The method includes the steps of: an acquisition step of acquiring a plurality of images, the images being images of a surface of a metal strip plate in a steady state in which rolling is performed by a rolling mill, i.e., images of the metal strip plate captured by a camera at a predetermined position under different lighting conditions or shooting conditions; a selection step of selecting, from each of the images acquired in the acquisition step, a reference pixel or a reference pixel range from a portion of the image other than a portion in which the metal strip plate is reflected, i.e., a portion other than a reflected light region in which reflected light is stronger than in other portions; a determination step of determining, from the plurality of images acquired in the acquisition step, a first image in which a reference pixel or a reference pixel range in which an average of R, G, and B values of the reference pixel or the reference pixel range has a maximum luminance; a calculation step of calculating a respective luminance difference between R, G, and B values of the reference pixel of the first image and R, G, and B values of a reference pixel of a region identical to the reference pixel of the first image in all other images other than the first image, or calculating a respective luminance difference between an average of R, G, and B values of the reference pixel range of the first image and an average of R, G, and B values of a reference pixel range of a region identical to the reference pixel range of the first image in all other images other than the first image; and a correction step of adding or subtracting an absolute value of the respective luminance difference to or from each of R, G, and B values of all pixel points of all other images so that each of R, G, and B values of a reference pixel of all other images coincides with R, G, and B values of the reference pixel of the first image, or so that each of R, G, and B values of an average of a reference pixel range of all other images coincides with R, G, and B values of an average of the reference pixel range of the first image.

2. The image correction method according to claim 1, wherein the reflected light region is a region in which a maximum value of luminance of R, G, and B values of the portion in which the metal strip plate is reflected is 0.9 times or more.

3. The image correction method according to claim 1 or 2, wherein the plurality of images acquired in the acquisition step are adjusted so that the luminance difference between an average of R values of the reference pixel or the reference pixel range of the first image is 51 or less.

4. The image correction method according to claim 1 or 2, wherein in the correction step, when a value of R, G, or B after the addition or subtraction of the absolute value of the luminance difference exceeds 255, the value of the corresponding R, G, or B is corrected to 255.

5. An abnormality detection method characterized by comprising: The method includes: the image correction method according to any one of claims 1 to 4; a step of calculating an abnormality detection judgment threshold value at the time of rolling using the first image; and a step of A step of detecting an abnormality of the surface of the metal strip based on the threshold value and an image obtained by a correction method according to any one of claims 1 to 4.

6. An image correction device characterized by comprising: Possessing: An acquisition unit that acquires a plurality of images, the images being images that include the surface of a metal strip in a steady state in which rolling is performed by a rolling mill in at least a portion thereof, that is, images that are captured by one camera at a prescribed position under different lighting conditions or shooting conditions and that represent the metal strip; A selection unit that selects a reference pixel or a reference pixel range from a portion of each of the images acquired by the acquisition unit that represents the metal strip, that is, a portion other than a reflected light region that is a region in which reflected light is stronger than in other portions, in the region that represents the metal strip; A determination unit that determines, from among the plurality of images acquired by the acquisition unit, an image in which the brightness of one of the average R value, G value, and B value of the reference pixel or the reference pixel range included in the image becomes the maximum brightness, as a first image; A calculation unit that calculates each of the brightness differences between the R value, G value, and B value of the reference pixel of the first image and the R value, G value, and B value of the reference pixel of the same region as the reference pixel of the first image in all of the other images other than the first image, or that calculates each of the brightness differences between the average R value, G value, and B value of the reference pixel range of the first image and the average R value, G value, and B value of the reference pixel range of the same region as the reference pixel range of the first image in all of the other images other than the first image; and A correction unit that adds or subtracts the absolute value of each of the brightness differences to or from each of the R value, G value, and B value of all of the pixels in all of the other images, so that each of the R value, G value, and B value of the reference pixel of the other images matches the R value, G value, and B value of the reference pixel of the first image, or so that each of the average R value, G value, and B value of the reference pixel range of the other images matches the average R value, G value, and B value of the reference pixel range of the first image.

7. The image correction device according to claim 6, wherein the reflected light region is a region in which the maximum value of the brightness of the R value, G value, and B value of the portion that represents the metal strip is 0.9 times or more.

8. The image correction device according to claim 6 or 7, wherein the plurality of images acquired by the acquisition unit are adjusted so that the brightness difference between the average R value of the reference pixel or the reference pixel range of the first image is 51 or less.

9. The image correction device according to claim 6 or 7, wherein the correction unit corrects the value of the corresponding R value, G value, and B value to 255 when the value of the R value, G value, and B value after the addition or subtraction of the absolute value of the brightness difference exceeds 255.

10. An abnormality detection device characterized by comprising: Possessing: The image correction device according to any one of claims 6 to 9; a threshold value operation section that calculates a rolling abnormality detection judgment threshold value using the first image; and an abnormality detection section that detects an abnormality of the surface of the metal strip based on the threshold value and the image acquired by the acquisition section and corrected by the image correction device according to any one of claims 6 to 9.

Images