Contour detection device and shape measurement device

Abstract

A contour detection device disclosed herein comprises: a camera for imaging a material to be rolled; and a processing circuit for processing a video captured by the camera. The processing circuit is configured to execute: dividing the video into a plurality of frames; identifying a plurality of target frames, in which the tip of the material to be rolled is present, among the plurality of frames; detecting a coordinate group representing the position of the material to be rolled in each target frame; calculating a coordinate conversion matrix for converting the coordinate group of each of the target frames into a preset reference coordinate group; obtaining one composite image from the plurality of target frames converted using the coordinate conversion matrix; and detecting the contour of the material to be rolled from the luminance of the composite image.

Description

Contour detection device and shape measurement device

[0001] The present disclosure relates to a contour detection device and a shape measurement device, and more particularly to a contour detection device that detects the contour of a rolled material transported in one direction on a hot rolling line, and a shape measurement device that measures the shape of the rolled material using the contour detected by the contour detection device.

[0002] There are shape measuring devices that measure shapes such as the warpage shape, camber, and flatness of a rolled material that has been rolled in a hot rolling line. Taking the warpage shape as an example, the following Patent Documents 1 to 3 disclose warpage shape measuring devices as shape measuring devices.

[0003] In Patent Document 1, a camera is installed on the same horizontal plane as the material to be rolled and at a position perpendicular to the widthwise center line (hereinafter also referred to as the "mill line") of the conveying table. This camera is used to capture an image of the widthwise end of the material to be rolled from directly beside it, the captured image is binarized, and the outline of the material to be rolled is extracted from the binarized image.

[0004] In Patent Document 2 listed below, a camera is installed diagonally above the gap between the stands (housings) of a finishing rolling mill so as to cover the gap within its field of view. The image captured by this camera is divided at a predetermined pitch along the rolling direction (longitudinal direction), the shape of the widthwise edge portion of the strip in each divided image is approximated by a quadratic equation, and the amount of warpage is quantified by a curvature.

[0005] In Patent Document 3, a first camera is installed to capture an image of the rolled material from the side direction, and a second camera is installed to capture an image of the rolled material from the conveyance direction. The edge shape extracted from the image captured by the first camera is corrected based on the crop shape extracted from the image captured by the second camera.

[0006] Japanese Patent Publication No. 1-285730 Japanese Patent No. 6828730 Japanese Patent No. 3724720

[0007] Incidentally, in order to measure the warpage shape with high accuracy, it is necessary to acquire a clear image. However, in a hot rolling line, the roller table is heated by the rolled material heated in the heating furnace, and high-pressure water is sprayed onto the heated roller table from a roll cooler or a descaler. This generates water vapor and causes water to accumulate on the rolled material. In addition, light reflected by side guides provided on both sides of the roller table in the width direction is irradiated onto the rolled material. If this water vapor, accumulated water, or reflected light appears in the image as disturbance, the acquired image becomes unclear, and as a result, the contour of the rolled material cannot be detected with high accuracy.

[0008] Since water vapor tends to stagnate for long periods of time, fans are sometimes installed to physically blow away the water vapor, but in low-temperature environments such as winter, a large amount of water vapor is generated, and there are limits to the fan's capacity.

[0009] The present disclosure has been made to solve the above-mentioned problems. An object of the present disclosure is to provide a contour detection device that can acquire a clear image suitable for detecting the contour of a rolled material even when disturbances such as water vapor occur. Another object of the present disclosure is to provide a shape measurement device that can accurately measure the shape of a rolled material from the contour detected from a composite image by the contour detection device.

[0010] The first aspect relates to a contour detection device that detects the contour of a rolled material being transported in one direction on a hot rolling line. The contour detection device includes a camera that captures images of the rolled material and a processing circuit that processes the images captured by the camera. The processing circuit is configured to divide the images into a plurality of frames, identify a plurality of target frames from the plurality of frames in which the leading edge of the rolled material is present, detect a set of coordinates representing the position of the rolled material in each of the target frames, calculate a coordinate transformation matrix that transforms the set of coordinates of each target frame into a set of predetermined reference coordinates, obtain a single composite image from the plurality of target frames transformed using the coordinate transformation matrix, and detect the contour of the rolled material from the brightness of the composite image.

[0011] The second aspect has the same features as the first aspect, but further includes the following: calculating the coordinate transformation matrix includes setting a set of coordinates of a reference frame selected from a plurality of target frames as a set of reference coordinates.

[0012] The third aspect has the same features as the first aspect, but further includes the following: detecting a coordinate group includes detecting a coordinate group including coordinates of four points for each target frame, and calculating a coordinate transformation matrix includes calculating a perspective transformation matrix as the coordinate transformation matrix.

[0013] A fourth aspect relates to a shape measurement device that measures the shape of a rolled material being transported in one direction on a hot rolling line. The shape measurement device includes the contour detection device according to any one of the first to third aspects and a shape measurement circuit. The shape measurement circuit is configured to measure the shape of the rolled material based on the contour of the rolled material detected from the composite image by the contour detection device.

[0014] The fifth aspect has the same features as the fourth aspect, but further includes the following: the shape measurement circuit is configured to measure the camber direction and camber height of the leading end of the rolled material.

[0015] The sixth aspect has the same features as the fifth aspect, but further includes the following: the shape measurement circuit is configured to obtain a monochrome image by performing a binarization process on the composite image, and to calculate the areas of white pixels in one or more regions of interest set in the monochrome image. Measuring the warp direction and warp height is performed when the areas of white pixels in the one or more regions of interest satisfy a determination criterion set for each region of interest.

[0016] According to the contour detection device of the present disclosure, multiple target frames in which the leading edge of the rolled material is present are combined using a coordinate transformation matrix. Even if a disturbance such as water vapor occurs, the disturbance does not exist in the same location in the multiple target frames, so the disturbance is canceled out when the multiple target frames are combined. As a result, the combined image obtained from the multiple target frames is clear, and the contour of the rolled material can be accurately detected from the clear combined image.

[0017] Furthermore, according to the shape measurement device of the present disclosure, the shape of the rolled material can be measured with high accuracy based on the contour of the rolled material detected from the clear composite image by the contour detection device. The present disclosure is particularly suitable for use in measuring the warpage direction and warpage height of the leading edge of the rolled material.

[0018] FIG. 1 is a schematic diagram showing a configuration of a warp measurement device equipped with a contour detection device according to an embodiment. FIG. 2 is a diagram showing an example of the hardware configuration of a processing circuit. FIG. 3 is a flowchart showing the flow of processing executed by the processing circuit. FIG. 4 is a diagram for explaining an example of a method for specifying a target frame. FIG. 5 is a diagram for explaining an example of a method for detecting the leading edge of a rolled material. FIG. 6 is a diagram for explaining an example of a method for defining a plane of the rolled material in each target frame. FIG. 7 is a diagram for explaining an example of a method for defining a plane of the rolled material in each target frame. FIG. 8 is a diagram for explaining an example of a method for detecting the contour of a rolled material from a composite image. FIG. 9 is a diagram for explaining an example of a method for quantifying a warp height from a contour. FIG. 10 is a diagram for explaining an example of a method for quantifying a warp height from a contour. FIG. 11 is a diagram for explaining perspective transformation of a contour. (a) and (b) are diagrams showing an example of a method for determining a warp direction and a method for calculating a warp height. FIG. 12 is a diagram showing an example of a method for compensating for plate width fluctuations occurring at the leading edge of a rolled material. FIG. 13 is a diagram showing an example of a method for calibrating a vanishing point using a calibration material. 10A and 10B are diagrams for explaining the effects of the present embodiment. 10B are diagrams showing other application examples of the present disclosure.

[0019] Hereinafter, with reference to the drawings, an embodiment of the present disclosure will be described taking as an example a case where the contour of a rolled material rolled by a roughing mill in a hot rolling line is detected and the warpage shape of the rolled material is measured from the detected contour. That is, an embodiment of a shape measurement device will be described taking as an example a warpage measurement device. Note that elements common to each drawing are assigned the same reference numerals and redundant explanations will be omitted.

[0020] 1 is a schematic diagram showing the configuration of a warp measurement device equipped with a contour detection device according to an embodiment. The warp measurement device 1 includes a contour detection device 2 and a shape measurement circuit (described later). The contour detection device 2 detects the contour of a rolled material Rm transported in one direction along a hot rolling line RL. The contour detection device 2 detects the contour of the rolled material Rm rolled by a roughing mill 3 on the hot rolling line RL.

[0021] The contour detection device 2 includes a camera 21 as an imaging means, a processing circuit 22, and a tracking unit 23. The processing circuit 22 also serves as the "shape measurement circuit" of the warpage measurement device 1, but a shape measurement circuit having the same configuration as the processing circuit 22 can also be provided separately.

[0022] As the camera 21, for example, a network camera with a resolution of 1280 x 720 pixels can be suitably used. The shooting mode of the camera 21 can be set to color, the frame rate to 30, the exposure time to a fixed value, and the aperture value to automatic mode. The resolution can be set according to the processing capacity of the processing circuit 22. The frame rate is the number of frames contained in one second. The exposure time and aperture value can be set in advance according to the predicted temperature of the rolled material Rm. Furthermore, the camera 21 is not limited to a network camera, and can also be used, for example, in combination with a machine vision camera or an infrared camera and a visible light cut filter. These cameras are well known, so further explanation will be omitted.

[0023] The camera 21 is installed so as to be able to capture an image of the material Rm being transported on the roller table 4 from obliquely above. For example, the camera 21 is tilted horizontally by an angle θ1 with respect to a line Lv that is perpendicular on a horizontal plane to the mill center line Lm, which is also the transport direction of the material Rm, and is tilted upward by an angle θ2 with respect to the horizontal plane, with a point O on the mill center line Lm as the center of view, and the distance from point O to the camera 21 is L. ｃａｍ The position of the point O and the angles θ1 and θ2 are preferably set so that there is no obstacle between the camera 21 and the rolled material Rm and that the position is as close as possible to the roughing mill 3. ｃａｍIt is desirable to set the distance L so that the field of view of the camera 21 can be secured to be about 5 m to 10 m in the direction of travel of the rolled material Rm. ｃａｍ can be changed depending on the lens (not shown) of the camera 21. The image captured by the camera 21 is input to a processing circuit 22.

[0024] The tracking unit 23 estimates the distance traveled by the tip of the rolled material Rm from the roughing mill 3 using the rolling load detected by the rolling load detection unit 31 of the roughing mill 3, the forward advance rate predicted from the rotational speed of the main machine 33 that rotates the rolling rolls 32, the rotational speed of the roller table 4, and the amount of reduction of the rolled material Rm.

[0025] The processing circuitry 22 processes the images captured by the camera 21. The processing circuitry 22 processes the images captured between a predetermined start timing and an end timing, among the images input from the camera 21. The start and end timings of the image capture are specified by signals from the tracking unit 23, which specify the timings when the leading end position is at a predetermined transport position. The tracking unit 23 can be configured as part of the processing circuitry 22.

[0026] FIG. 2 is a diagram illustrating an example of the hardware configuration of a processing circuit. As shown in FIG. 2, the processing circuit 22 may include dedicated hardware 22a, a processor 22b, and a memory 22c. At least a portion of the processing circuit 22 may be at least one dedicated hardware 22a. In this case, the processing circuit 22 may be, for example, a single circuit, a composite circuit, a programmed processor, a parallel programmed processor, an ASIC, an FPGA, or a combination thereof. The processing circuit 22 may include at least one processor 22b and at least one memory 22c. In this case, each function of the contour detection device 2 is realized by software, firmware, or a combination of software and firmware. The software and firmware are written as programs and stored in the memory 22c. The processor 22b realizes the functions of each part of the contour detection device 2 by reading and executing the programs stored in the memory 22c. The processor 22b is also called a CPU (Central Processing Unit), central processing unit, processing device, arithmetic unit, microprocessor, microcomputer, or DSP. The memory 22c may be, for example, a non-volatile or volatile semiconductor memory such as a RAM, a ROM, a flash memory, an EPROM, an EEPROM, etc. In this way, the processing circuitry 22 can realize each function of the contour detection device 2 by hardware, software, firmware, or a combination of these.

[0027] Processing circuitry 22 is configured to perform the following: Figure 3 is a flow chart illustrating the processing performed by processing circuitry 22.

[0028] First, the processing circuit 22 divides the video image input from the camera 21 into a plurality of frames (step S1). The plurality of frames are, for example, color images, but are not limited to this and may be images obtained by converting a color image into grayscale.

[0029] The processing circuitry 22 applies a differential filter to each of the multiple frames. The timing of applying the differential filter is not limited to this. The differential filter may be applied to a target frame identified in step S2 (described later), or to a composite image obtained in step S5 (described later). A Laplacian filter, which performs second-order differentiation, is preferably used as the differential filter. However, the differential filter is not limited to this. A Sobel filter, which performs first-order differentiation, or a Scharr filter, which performs third-order differentiation, may also be used. Using such a differential filter results in an edge image in which areas where brightness suddenly changes are emphasized as edges. The frames to which the differential filter is applied are not limited to color images. A single or multiple RGB channels of a color image may be selected and applied. The differential filter may also be applied to an image obtained by converting a color image to grayscale. Alternatively, the differential filter may be applied to a single or multiple RGB channels. In the case of multiple channels, the filtered images may be combined and then converted to grayscale.

[0030] Next, the processing circuit 22 identifies a target frame from among the multiple frames (edge ​​images) (step S2). In step S2, it is determined whether the multiple frames correspond to the target frame. A target frame is a frame to be combined (combining process), and the leading edge of the rolled material Rm is located within a predetermined range of the frame. FIG. 4 is a diagram for explaining an example of a method for identifying a target frame. As shown in FIG. 4, in frame (hereinafter also referred to as "image") 5, a region of interest (ROI) 51 is set upstream of the mill center line Lm, and a region of interest 52 is set downstream of the mill center line Lm. Frame 5 is binarized into maximum and minimum brightness values. If the number of pixels with the maximum brightness in region of interest 51 is equal to or greater than a threshold value and the number of pixels with the maximum brightness in region of interest 52 is equal to or less than a threshold value, the frame can be determined to be the target frame. In this case, it is determined that the leading edge of the rolled material Rm is located between the two regions of interest 51 and 52. The shapes of the regions of interest 51 and 52 are, for example, rectangular (square in FIG. 4). As another method of identification, the tip position detected when detecting the coordinate group by the method described later with reference to Fig. 5 may be used, and the frame may be determined to be the target frame if the tip position is located within a predetermined tip detection range 53. As will be described later, reference numeral 54 in Fig. 5 denotes a moving window that slides within the tip detection range 53 along the mill center line Lm.

[0031] The processing circuitry 22 executes steps S3 to S8 on a plurality of target frames. Steps S3 to S5 are a synthesis process for synthesizing a plurality of target frames to obtain a single synthesized image. Steps S6 to S8 are well-known conventional processes that have been conventionally performed.

[0032] In step S3, the processing circuit 22 detects a group of coordinates as feature quantities of the rolled material Rm in each target frame. In this step S3, a group of coordinates of the end points of a rectangle that exists on the same horizontal plane as the leading end position P defined based on the leading end position P of the rolled material Rm is detected as feature quantities of the rolled material Rm.

[0033] FIG. 5 is a diagram illustrating an example of a method for detecting the leading edge of the rolled material Rm. As shown in FIG. 5, a leading edge detection range 53 is defined within the target frame 5 corresponding to the field of view of the camera 21, and a moving window 54 is defined that slides within the leading edge detection range 53 along the mill center line Lm. The leading edge detection range 53 and the moving window 54 are set to have constant dimensions in the strip width direction. Furthermore, regardless of the sliding position within the leading edge detection range 53, the long side of the moving window 54 is parallel to the strip width direction of the rolling equipment, and the short side is parallel to the longitudinal direction of the rolling equipment. While the moving window 54 is sliding, the average brightness value within the moving window 54 is calculated, and the sliding position of the moving window 54 and the average brightness value are associated and stored in memory 22c. Of the average brightness values ​​stored in memory 22c, the sliding position at which the average brightness value is maximized is detected as the leading edge position.

[0034] 6, 7, and 8 are diagrams for explaining an example of a method for defining the plane of the rolled material Rm in each target frame. As shown in Fig. 6, 7, and 8, the plane of the rolled material Rm in the target frame 5 is defined by the vertices of a rectangle (four points a to d), and a coordinate group including the coordinates of these four points is detected as a feature.

[0035] In the example shown in Figure 6, using the leading edge position P on the mill center line Lm as a reference, points a and b are defined as points that are shifted a distance Lw from the leading edge position P to the positive side (the rear side in the figure) or the negative side (the front side in the figure) in the width direction, respectively. Points that are shifted a distance Ll from points a and b toward the tail end in the longitudinal direction are defined as points d and c, respectively. Note that points that are shifted a distance Ll from points a and b toward the leading edge in the longitudinal direction may also be defined as points d and c, respectively.

[0036] In the example shown in Figure 7, using the leading edge position P on the mill center line Lm as a reference, edge detection is performed on a straight line extending on both sides in the width direction from point Q, which is moved a distance Ll from the leading edge position P toward the longitudinal tail end. The points at the width ends of the rolled material Rm obtained by edge detection are defined as points c and d, respectively. Then, points moved a distance Ll from points c and d toward the leading edge in the longitudinal direction (outside the rolled material Rm) are defined as points b and a, respectively.

[0037] In the example shown in Figure 8, point c, which is the width end of the rolled material Rm, cannot be detected because point c or point d (point c in this example) overlaps with a side guide 41, which is part of the rolling equipment. In this case, the coordinates of point c are detected using the following method. That is, point c is defined as a point that is moved a distance Lb in the width direction from point d, which is the width end of the rolled material Rm that has been correctly detected. This movement distance Lb may be the calculated or measured strip width of the rolled material Rm, or it may be set to the maximum value of the length from point d to point c (corresponding to the strip width) or twice the length from point d to point Q, detected using a method similar to the example in Figure 7, by gradually increasing the distance Ll.

[0038] In step S4, the processing circuitry 22 calculates a coordinate transformation matrix. In step S4, a perspective transformation matrix is ​​calculated as the coordinate transformation matrix. The processing circuitry 22 first sets (selects) one reference frame 5r from among the multiple target frames 5, and sets the coordinate group (feature values) of four points (points a to d) in the one reference frame 5r as a reference coordinate group. Next, the processing circuitry 22 calculates a perspective transformation matrix M that transforms the coordinate group of four points (points a to d) in each target frame 5 other than the reference frame 5r into the reference coordinate group of four points (points a to d) in the reference frame 5r. Note that the reference coordinate group does not necessarily have to be set on the reference frame 5r, but may be set in advance on the target frame 5. In this case, in step S5 described below, the image after perspective transformation is not superimposed on the reference frame 5r, but rather the images after perspective transformation (target frames) are superimposed on each other to obtain a composite image 6.

[0039]

[0040] Here, i ｒ is the X-axis position (X coordinate) of the point in the reference frame 5r, j ｒ is the Y-axis position (Y coordinate) of the point in the reference frame 5r, i ｎ is the X-axis position (X coordinate) of the point in the target frame 5n, j ｎ is the Y-axis position (Y coordinate) of the point in the target frame 5n, h is a component of the perspective transformation matrix Mn, and k is the position of the point cloud (k=a, b, c, d).

[0041] In each target frame 5 other than the reference frame 5r, perspective transformation is performed using the calculated perspective transformation matrix M. At this time, the image I before perspective transformation is ｉｎ Each pixel of the image I after perspective transformation is calculated using each component of the perspective transformation matrix M of the target frame 5 as shown in the following equation (2): ｔｒａｎｓ The perspective transformation matrix M is used to transform the image I ｔｒａｎｓ Each pixel of the image I before perspective transformation ｉｎ It is a matrix that indicates which pixel each corresponds to.

[0042] In the above equation (2), i is the image I before perspective transformation. ｉｎ , j is the X-axis position of the pixel in the image I before perspective transformation. ｉｎ is the Y-axis position of the pixel in the image, m is the channel of the image, and h is a component of the perspective transformation matrix M. However, the image I after perspective transformation ｔｒａｎｓ The image I before perspective transformation is ｉｎ If there is no pixel in the image I ｔｒａｎｓ The brightness of the pixel is set to 0.

[0043] Here, the consistency of the leading edge position P detected in each target frame 5 may be verified using conditions, and target frames 5 that do not satisfy the conditions may be excluded from the compositing target. For example, a statistical value is calculated from the relationship between the leading edge position P of each target frame 5 and the average brightness value within the moving window 54, and target frames 5 whose statistical value deviates from the distribution of all target frames 5 are excluded from the compositing target. Another method for checking the consistency based on the leading edge position P of each target frame 5 is as follows. That is, if the recording start timing and end timing are adjusted so that the traveling direction of the rolled material Rm in the video is constant, the detected leading edge position P will travel in the same direction between target frames 5. Based on this relationship, if the leading edge position P of a target frame 5 is not located between the leading edge positions P of the target frames 5 before and after it, the target frame 5 is excluded from the compositing target. Alternatively, the actual values ​​of the tracking unit 23 from the start to the end of recording are collected and compared with the leading edge position P of each target frame 5. If there is a large difference, the target frame 5 is excluded from the compositing target.

[0044] In step S5, the processing circuitry 22 superimposes all of the images of the target frames that have been perspective transformed using the perspective transformation matrix M onto the reference frame 5r, thereby obtaining a composite image 6. The luminance of each pixel of the composite image 6 is expressed by the following equation (3):

[0045] In the above formula (3), I ｏｕｔ is the composite image 6, i is the X-axis position of the pixel, j is the Y-axis position of the pixel, m is the channel of the image, I ｔｒａｎｓ is the image after perspective transformation (target frame 5), n is the serial number of the target frame, and N is the total number of target frames.

[0046] Steps S6 to S8 are conventional processes for acquiring information used in step S13, which will be described later. In step S6, the processing circuit 22 detects contours of areas with large changes in brightness (boundaries between maximum and minimum brightness) from an edge image obtained by applying a differential filter to each target frame. Based on the detected contours, the processing circuit 22 determines the warp direction (upward or downward) (step S7) and calculates the warp height (step S8). The calculated warp height for each target frame is temporarily stored in the memory 22c and read out during the processing of step S13, which will be described later. The conventional processing of steps S6 to S8 and the processing of step S13, which measures warp using statistical processing, are publicly known, including the disclosures of Patent Documents 1 to 3, and therefore will not be described in detail here. Once the processing of step S8 has been completed for all target frames, the processing proceeds to step S9.

[0047] However, if the rolled material Rm has a downward bow and the bow height is large, the leading end of the rolled material Rm may collide with the roller table 4 and bounce up and down. When such a bound occurs, an afterimage is generated at the leading end of the rolled material Rm in the composite image 6. This reduces the brightness of the leading end in the composite image 6, making it impossible to accurately detect the outline of the leading end of the rolled material Rm, which may result in an erroneous measurement of an upward bow. As a result of extensive research, the inventors have discovered that when an afterimage is generated due to the bound, the bouncing portion also appears white in the black and white image obtained by performing a binarization process on the composite image 6, resulting in an increase in white pixels.

[0048] Based on the above findings, in step S9, a black-and-white image (binarized image) is obtained by performing a binarization process on the composite image. Next, the area of ​​white pixels in the black-and-white image is calculated (step S10). In step S10, the area of ​​white pixels in at least one (one or more) region of interest (e.g., regions of interest 51 and 52 shown in FIG. 4 ) previously set in the black-and-white image (i.e., the composite image) is calculated. Here, at least one region of interest can be set to include a portion affected by bounding, taking into account, for example, past performance and a reference coordinate group. When multiple regions of interest are set, the multiple regions of interest may be set adjacent to each other or spaced apart.

[0049] Next, it is determined whether the area of ​​white pixels calculated in step S10 satisfies a criterion (step S11). When multiple regions of interest are set, a criterion can be set for each region of interest. Examples of the criterion include exceeding a reference value or being equal to or less than a reference value. The criterion, including the reference value, is set for each region of interest. For example, when two regions of interest are set, if the area of ​​white pixels in the first region of interest exceeds a reference value and the area of ​​white pixels in the second region of interest is equal to or less than the reference value, it can be determined that the criterion is satisfied. In this way, the criterion can be set arbitrarily for each region of interest.

[0050] If it is determined in step S11 that the judgment criterion is met, it is determined that no bounding has occurred, and the process proceeds to step S12. On the other hand, if it is determined that the judgment criterion is not met, it is determined that bounding has occurred, and the process proceeds to step S13. In step S13, the warpage is measured by statistically processing the warpage directions and warpage heights of multiple target frames that have not been combined, using a method similar to that used in the conventional example. In the statistical processing, for example, the median of the warpage heights obtained for multiple target frames is obtained, but this is not limited thereto, and the maximum value may be obtained, or the median of the results (warpage heights) excluding the maximum and minimum values ​​may be obtained.

[0051] In step S12, the processing circuitry 22 detects the contour Ed of the rolled material Rm from the composite image 6. FIG. 9 is a diagram for explaining an example of a method for detecting the contour Ed of the rolled material Rm from the composite image 6. First, the leading edge position (leading edge point) P of the rolled material Rm in the composite image 6 is detected. Then, a region of interest (ROI) 55 is defined, having a length Lsw (mm) in the strip width direction from the leading edge point P on the mill center line Lm and a length Lsl (mm) in the longitudinal direction. Then, a moving window 56 is defined, which moves longitudinally within the ROI 55 and has the same length Lsw (mm) in the strip width direction as the ROI 55, but has an arbitrary length Lml (mm) in the longitudinal direction that is smaller than Lsl. Furthermore, a moving window 57 is defined, which moves longitudinally within the moving window 56 and has the same length Lml (mm) in the longitudinal direction as the moving window 56, but has an arbitrary length Lmw (mm) in the strip width direction that is smaller than Lsw. At each longitudinal position of the moving window 56, the average brightness value within the moving window 57 when the moving window 57 is moved in the sheet width direction is stored in the memory 22c in association with the moving position. At each longitudinal position of the moving window 56, the center point C of the moving window 57 when the average brightness value within the moving window 57 is maximum is detected as the contour Ed of the rolled material Rm.

[0052] Here, the contour Ed is detected using a composite image 6 obtained by combining edge images of each target frame 5 through perspective transformation, but this is not limiting and similar results can be obtained using other processes. For example, a binarized image can be obtained by performing binarization processing on the composite image 6, and the center position in the plate width direction of the pixel with the maximum brightness at each longitudinal position within the region of interest 55 can be detected as the contour Ed of the rolled material Rm. Note that a noise removal filter may be applied to the composite image 6 as necessary.

[0053] Next, with reference to Figures 10 to 14, the processing circuit 22 serving as a shape measurement circuit quantifies the warpage height from the detected contour Ed. First, as shown in Figure 10, a contour Ed1 on the rear side in the width direction is extracted from the contour Ed of the rolled material Rm. The extracted contour Ed1 is subjected to perspective transformation, and the warpage direction (upward or downward warpage) is determined from the characteristics of the contour Ed2 after perspective transformation. Then, the warpage height is quantified based on the conditions for each warpage direction. Below, the perspective transformation of the contour Ed1, determination of the warpage direction, and calculation of the warpage height, which are executed by the processing circuit 22, will be described.

[0054] (Perspective Conversion of Contour) To convert the unit of warp height [pix] on the image to [mm], an image of the rolled material Rm viewed from the side is obtained using the method described below. That is, the processing circuit 22 defines the plane on which the widthwise rearward contour Ed1 exists, i.e., the plane on which the warp height is calculated, using the following method. As shown in FIG. 11 , the reference line Lr is defined as an extension of the non-warped portion of the contour Ed1. The intersection of the reference line Lr and a predetermined line t3-t4 (described below) is detected as point r1. The intersection of the reference line Lr and a predetermined line t1-t2 (described below) is detected as point r2. Then, as shown in FIG. 12 , the intersection of a line connecting point r1 and vanishing point VP3 (described below) with a predetermined line t8-t7 is detected as point r4. Then, the intersection of the straight line connecting point r2 and vanishing point VP3 with the predetermined straight line t5-t6 is detected as point r3. The points r1, r2, r3, and r4 detected by the above process define a plane Pn1. The short sides of the plane pn1 extend in the thickness direction, and the long sides extend in the longitudinal direction. Then, a matrix M is used to perspectively transform the plane Pn1 into a rectangular (longitudinal) plane Pn2 (plane Pn2 defined by points r1', r2', r3', and r4') with predetermined side lengths as shown in FIG. ｍｅａｓ is calculated. As shown in FIG. 13 , the calibration material Cm has a known thickness direction dimension Tc [mm] and longitudinal direction dimension Lc [mm]. From these dimensions Tc and Lc and the desired pixel size conversion coefficient Sdes [mm / pix], when the short side of the rectangle is set to Tc / Sdes [pix] and the long side is set to Lc / Sdes [pix], the line in the thickness direction and the line in the longitudinal direction are perpendicular to each other on the plane Pn2 after perspective transformation, so that the desired pixel size conversion coefficient Sdes [mm / pix] is obtained at any position on the image. In other words, the unit of the warp height can be converted from [pix] to [mm] using one pixel size conversion coefficient Sdes.

[0055] (Determination of Warp Direction) Next, the processing circuitry 22 converts the detected points constituting the contour Ed1 on the far side in the width direction into M ｍｅａｓBy performing perspective transformation using the formula (1), a contour Ed2 after perspective transformation as shown in FIG. 14 is obtained. FIG. 14 is a diagram showing an example of a method for determining the warpage direction. Among the points on the contour Ed2 after perspective transformation, the highest point p in the Y-axis direction (the up-down direction in the figure) is defined as the peak point, and the point q at the tip of the contour Ed2 is defined as the leading end point. In this case, if the longitudinal distance d from the peak point p to the leading end point q is smaller than a preset threshold, it is determined to be an upward warpage (see FIG. 14( a)), and if it is greater than the threshold, it is determined to be a downward warpage (see FIG. 14( b)). The threshold can be set appropriately based on past performance.

[0056] In this way, the warpage direction is determined based on the longitudinal distance d from point p to point q of the contour Ed2 on the rear side in the width direction near the front end of the rolled material Rm, but this is not limiting. For example, the warpage direction can be determined by performing a similar process on the contour on the front side in the width direction near the front end of the rolled material Rm. In this case, the warpage height calculation described below can be performed only when the warpage directions on the front side and the rear side in the width direction match. When the warpage directions determined on the rear side and the front side in the width direction do not match, the warpage height calculation is performed using the method described below.

[0057] (Calculation of Warpage Height) In step S12, the processing circuit 22 calculates the warpage height based on the contour Ed2 on the rear side in the width direction. In the case of upward warpage, as shown in Fig. 14(a), the lowest point in the Y-axis direction of the contour Ed2 of the part without warpage is defined as the lowest point r, and the difference in height h in the thickness direction between the peak point p and the lowest point r is calculated as the warpage height. In the case of downward warpage, as shown in Fig. 14(b), the difference in height h in the thickness direction between the peak point p and the tip point q is calculated as the warpage height. The height of upward warpage is a positive value, and the height of downward warpage is a negative value.

[0058] It is of course possible to perform the above-mentioned image synthesis and warpage measurement after one rolling pass by the roughing mill 3 is completed, and to apply control in accordance with the measured warpage (for example, change in the rotation speed of the upper and lower rolling rollers 32) to the next rolling pass, but by performing the image synthesis and warpage measurement during a rolling pass, it is also possible to apply them during the current rolling pass, thereby realizing real-time performance. When performing the image synthesis and warpage measurement during the current rolling pass, if the predetermined number of target frames 5 cannot be obtained, it is also possible to configure so that the data of past rolling passes is read from the memory 22c to obtain the synthesized image 6.

[0059] (Method for Compensating for Strip Width Variation) With reference to FIG. 15 , a method for quantifying the warpage height when strip width variation occurs at the leading end of the rolled material Rm will be described. When strip width variation occurs at the leading end of the rolled material Rm, if the above-described warpage direction determination is performed on the contour Ed2 on the rear side in the strip width direction and the contour on the front side in the width direction, respectively, the results of the two warpage direction determinations may not match. For example, as shown in FIG. 15 , when the width near the leading end of the rolled material Rm is narrow, downward warpage may be erroneously detected based on the contour Ed2 on the rear side in the strip width direction and the distance da, while excessive upward warpage may be erroneously detected based on the contour Ed3 on the front side in the width direction and the distance db. Therefore, the following method can be adopted to compensate for the width variation at the leading end of the rolled material Rm and accurately measure the warpage height. That is, line segments extending in the width direction on the rolled material Rm are detected at each position in the longitudinal direction (hereinafter referred to as "each longitudinal position"), and points Pc which are the widthwise centers of each line segment are detected as the center point, and a center line La is obtained by connecting the center points Pc at each longitudinal position. This center line La is used instead of the contours Ed2, Ed3 on both sides in the width direction to perform the above-mentioned perspective transformation, judgment of the warpage direction, and calculation of the warpage height. This makes it possible to compensate for width variations at the leading edge of the rolled material Rm, and to measure the warpage height with high accuracy. Note that this compensation method is based on the premise that width variations at the leading edge of the rolled material Rm occur approximately evenly on the rear and front sides in the width direction.

[0060] (Calibration Method) FIG. 16 shows an example of a method for calibrating vanishing points VP1, VP2, and VP3 using a calibration material Cm. The calibration material Cm has a known plate thickness Tc, plate width Wc, and length Lc. The calibration material Cm is positioned so that its widthwise center coincides with the mill center line Lm, and an image of the calibration material Cm is captured. Points t1 to t6 are detected on the captured image. The vanishing point VP1 is defined as the position where a line connecting points t1 and t2 intersects with a line connecting points t3 and t4. Similarly, the vanishing point VP2 is defined as the position where a line connecting points t1 and t4 intersects with a line connecting points t2 and t3. Similarly, the vanishing point VP3 is defined as the position where a line connecting points t1 and t5 intersects with a line connecting points t2 and t6. If the X-coordinate or Y-coordinate, or both, of these vanishing points VP1, VP2, and VP3 exceed a predetermined value significantly greater than the number of pixels in the frame, the vanishing point is deemed not to exist, and lines parallel to the above lines are defined as lines in each axial direction. The reference lines are the line connecting points t1 and t5 in the thickness direction, the line connecting points t1 and t2 in the width direction, and the line connecting points t1 and t4 in the longitudinal direction. Next, the intersection of lines (t7-VP1), (t5-VP2), and (t4-VP3) is detected as point t8. The coordinates of vanishing points VP1-VP3 and points t1-t8 on the image defined by the above operations are stored in memory 22c and are used to detect the above-mentioned coordinate group (feature values) and calculate the warp height.

[0061] As described above, according to this embodiment, a plurality of target frames 5 in which the tip of the rolled material Rm exists are synthesized using the coordinate transformation matrix M. As shown in FIG. 17, even if a disturbance such as steam or stagnant water occurs, the disturbance is transmitted to a plurality of target frames 5 (four in the figure). １ ~5 ４ Since the image does not exist in the same location in multiple target frames 5 １ ~5 ４ As a result, the disturbance is cancelled when combining the target frames 5 １ ~5 ４The composite image 6 obtained from this is clear, and the outline Ed of the rolled material Rm can be accurately detected from the clear composite image 6. Furthermore, the warpage of the rolled material Rm can be measured with high accuracy based on the outline Ed of the rolled material detected from the clear composite image 6.

[0062] Although the embodiments of the present disclosure have been described above, the present disclosure is not limited to the above embodiments and can be implemented in various modifications without departing from the spirit of the present disclosure. When the numbers, quantities, amounts, ranges, etc. of each element are mentioned in the above embodiments, the present invention is not limited to the mentioned numbers unless otherwise specified or clearly specified in principle. Furthermore, the structures, etc. described in the above embodiments are not necessarily essential to the present invention unless otherwise specified or clearly specified in principle.

[0063] In the above embodiment, an example has been described in which the warpage direction and warpage height of the rolled material Rm are measured, but the shape of the rolled material is not limited to warpage. The present disclosure can also be applied to measuring shapes that are measured based on the outline Ed of the rolled material Rm, such as the camber shown in Figure 18(a) and the wavy surface shape called flatness shown in Figure 18(b).

[0064] In the above embodiment, a group of coordinates of four points at the front end of the rolled material Rm was detected as a feature, but in cases where the rolled material Rm is short, a group of coordinates of four points at the tail end of the rolled material Rm can also be detected as a feature.

[0065] In the above embodiment, an example has been described in which a perspective transformation matrix M is calculated to perform perspective transformation on the target frame 5 so that the shape and size of the rolled material Rm in the reference frame 5r coincide with the shape and size of the rolled material Rm in the target frame 5, but this is not limited to this. When the camera 21 is installed far from the mill center line Lm, the size of the rolled material Rm in multiple target frames 5 is the same, and the rolled material Rm can be considered similar, i.e., when size changes in the rolled material Rm in multiple target frames 5 can be ignored, a composite image can be obtained by translation. In this case, the coordinates of three points are detected for each target frame 5, and an affine transformation matrix is ​​calculated as a coordinate transformation matrix. Because the affine transformation matrix does not need to include rotation, the affine transformation matrix can be expressed as a 3 × 2 matrix.

[0066] In the above embodiment, an example has been described in which the contour Ed of the rolled material Rm rolled by the roughing mill 3 is detected, but the present disclosure is not limited to this. The present disclosure can be applied to cases in which the contour of the rolled material Rm transported in one direction along the hot rolling line RL is detected, such as when detecting the contour of the rolled material Rm rolled by the finishing mill.

[0067] RL...hot rolling line, Rm...rolled material, 1...shape measurement device, 2...contour detection device, 21...camera, 22...processing circuit (shape measurement circuit), 23...tracking unit, 3...roughing mill, 31...rolling load detection unit, 32...rolling roll, 33...main machine, 4...roller table, 41...side guide, 5...object frame, 5r...reference frame, 6...synthetic image

Claims

1. A contour detection device for detecting the contour of a rolled material being transported in one direction on a hot rolling line, comprising: a camera for capturing images of the rolled material; and a processing circuit for processing the images captured by the camera, wherein the processing circuit is configured to: divide the image into a plurality of frames; identify a plurality of target frames from the plurality of frames in which the leading edge of the rolled material is present; detect a group of coordinates representing the position of the rolled material in each target frame; calculate a coordinate transformation matrix for converting the group of coordinates of each target frame into a predetermined group of reference coordinates; obtain a single composite image from the plurality of target frames transformed using the coordinate transformation matrix; and detect the contour of the rolled material from the brightness of the composite image.

2. A contour detection device according to claim 1, wherein calculating the coordinate transformation matrix includes setting the coordinate group of a reference frame selected from the plurality of target frames as the reference coordinate group.

3. A contour detection device according to claim 1, wherein detecting the coordinate group includes detecting the coordinate group including the coordinates of four points for each target frame, and calculating the coordinate transformation matrix includes calculating a perspective transformation matrix as the coordinate transformation matrix.

4. A shape measuring device for measuring the shape of a rolled material being transported in one direction on a hot rolling line, comprising: a contour detection device according to any one of claims 1 to 3; and a shape measurement circuit configured to measure the shape of the rolled material based on the contour of the rolled material detected from the composite image by the contour detection device.

5. A shape measuring device according to claim 4, wherein said shape measuring circuit is configured to measure the warpage direction and warpage height of the leading end of said rolled material.

6. A shape measurement device according to claim 5, wherein the shape measurement circuit is configured to obtain a black-and-white image by performing binarization processing on the composite image, and to calculate the area of ​​white pixels in one or more regions of interest set in the black-and-white image, and wherein measuring the warp direction and the warp height is performed when the area of ​​white pixels in the one or more regions of interest satisfies a judgment criterion set for each region of interest.

Images