Tubular belt conveyor tube expansion detecting and adjusting method and device

Abstract

The invention provides a tubular belt conveyor tube expansion detecting and adjusting method and device. A camera fixed at the upper end of a pipe belt machine is used for obtaining a real-time video, after preprocessing of perspective transformation, gray level conversion, bilateral filtering noise reduction and the like is conducted, edge detection is conducted by combining a self-adaptive multi-scale wavelet edge detection algorithm with a one-way edge detection operator, and an edge straight line is fitted through an LSD algorithm to calculate the pipe diameter. When pipe expansion is detected, the pipe diameter adjusting device is driven to adjust the pipe diameter, and closed-loop operation is formed. According to the method, the edge detection precision is improved by improving the wavelet edge detection method, the edge position of the pipe belt can be detected in real time, contact abrasion is avoided, the pipe expansion problem is solved in real time, and the labor cost and economic loss are reduced.

Description

Technical Field

[0001] The present invention relates to the control technology of pipe belt conveyors, and particularly to a method and device for detecting and adjusting the expansion of a tubular belt conveyor. Background Art

[0002] Compared with traditional belt conveyors, the tubular belt conveyor adopts a completely enclosed transportation method, which can effectively prevent material scattering, and can also adapt to complex terrains and spatial layouts. It has the advantages of high environmental protection, high space utilization rate, high transportation efficiency, etc., and has widely replaced traditional belts in the transportation of coal, ore, grain, etc.

[0003] The fault detection of the pipe belt conveyor can be carried out in the directions of tearing, deviation, etc. However, when tearing occurs, huge economic losses have already occurred. Before the pipe belt conveyor tears, problems such as pipe expansion often occur. The pipe expansion of the pipe belt conveyor caused by uneven material flow and abnormal conveyor belt tension is one of the important reasons for the tearing of the pipe belt conveyor. If the pipe expansion can be detected and effectively dredged at this stage, the economic loss can be greatly reduced. Therefore, the detection of pipe expansion of the pipe belt conveyor is crucial.

[0004] However, the traditional pipe expansion detection uses a mechanical trigger device, which has the disadvantages of too fast wear, low accuracy, and being greatly affected by environmental factors, and it is not easy to accurately detect the pipe expansion of the pipe belt conveyor. Therefore, it has become an urgent problem to improve the accuracy of pipe expansion detection of the pipe belt conveyor, reduce costs, and adjust the pipe expansion problem in real time through the method of combining machine vision and image processing. Summary of the Invention

[0005] The present invention aims at the deficiencies in the prior art and provides a method and device for detecting and adjusting the expansion of a tubular belt conveyor.

[0006] To achieve the above object, the present invention adopts the following technical solutions:

[0007] A method for detecting and adjusting the expansion of a tubular belt conveyor includes the following steps:

[0008] Taking pictures of the tubular belt conveyor;

[0009] Performing perspective transformation on the pictures of the tubular belt conveyor to convert the viewing angle of the pictures of the tubular belt conveyor to a top view angle;

[0010] Converting the colored pictures of the tubular belt conveyor into grayscale images, and performing denoising processing on the grayscale images by using bilateral filtering;

[0011] Performing edge detection on the denoised image by using an adaptive multi-scale wavelet edge detection algorithm and performing binary processing on the result to obtain a first edge picture;

[0012] Edge detection is performed on the denoised image using a Laplacian operator with increased horizontal direction weights to obtain a second edge picture;

[0013] Calculate the SSIM values of the first edge picture and the second edge picture, dynamically adjust the fusion weights based on the SSIM values, and perform weighted fusion on the first edge picture and the second edge picture;

[0014] Use the LSD algorithm to fit two best edge lines in the fused edge picture as the pipe belt edge lines;

[0015] Calculate the maximum distance between the pipe belt edge lines. If the maximum distance is greater than or equal to the threshold, it is determined as tube expansion;

[0016] Increase the pipe diameter after receiving the tube expansion signal.

[0017] To optimize the above technical solution, the specific measures taken also include:

[0018] Further, the perspective transformation of the tubular belt conveyor picture is specifically as follows:

[0019] Select four groups of idler vertices on both sides of the pipe belt as 4 points (x i , y i ) in the source image and 4 points (x i ', y i ) in the target image. Obtain the perspective transformation matrix H by solving the following linear equations:

[0020]

[0021] In the formula, h ij is the element in the i-th row and j-th column of the perspective transformation matrix,

[0022] Perform perspective transformation on the point (x, y) on the tubular belt conveyor picture through the following transformation formula:

[0023]

[0024] In the formula, (x’, y’) is the position of the new point after mapping.

[0025] Further, the denoising process of the grayscale image using bilateral filtering is specifically as follows:

[0026] For each pixel p in the grayscale image, its grayscale value I bf (p) is calculated by the following formula:

[0027]

[0028] Where: S is the neighborhood centered on pixel p, is the Gaussian kernel function in the spatial domain, used to measure the distance weight of pixels p and q in terms of spatial position. is the Gaussian kernel function in the gray value range domain, used to measure the difference weight of pixels p and q in terms of gray values. I(p) represents the original gray value of pixel q in the neighborhood S, and I(q) is the original gray value of pixel q in the neighborhood S, where q is any pixel in the neighborhood S of p that is not equal to p. is the normalization factor;

[0029] The expression of the Gaussian kernel function in the spatial domain is: where d = ||p - q|| is the Euclidean distance between pixels p and q, and s d is the standard deviation of the Gaussian kernel in the spatial domain;

[0030] The expression of the Gaussian kernel function in the gray value range domain is: where DI = |I(p) - I(q)| is the difference in gray values between pixels p and q, and s r is the standard deviation of the Gaussian kernel in the gray value range domain;

[0031] Calculate for each pixel q in the neighborhood S, then sum and divide by the normalization factor W p to obtain I bf (p). Calculate each pixel in the image according to this method to complete bilateral filtering denoising.

[0032] Furthermore, the specific process of edge detection of the denoised image by the adaptive multi-scale wavelet edge detection algorithm is as follows:

[0033] Select Symlets wavelet for three-layer wavelet decomposition. In each layer of decomposition, perform convolution operations on the image through the low-pass filter h and the high-pass filter g. For the horizontal direction:

[0034]

[0035] For the vertical direction:

[0036]

[0037] where, I i-1 is the image of the previous layer, s represents the spatial index of I i-1 in the horizontal direction, t represents the spatial index of I i-1 in the vertical direction, m represents the spatial index of the component obtained after convolution operation of I i in the horizontal direction, n represents the spatial index of the component obtained after convolution operation of I i in the vertical direction, LL i is the low-frequency approximation component of the i-th layer, LHi The value of the component representing the combination of the low frequency in the horizontal direction and the high frequency in the vertical direction in the i-th layer wavelet decomposition, HL i is the value of the component representing the combination of the high frequency in the horizontal direction and the low frequency in the vertical direction in the i-th layer wavelet decomposition, HH i Represents the value of the component where both the horizontal and vertical directions are high frequency in the i-th layer wavelet decomposition, N is the size of the image; after three-layer decomposition, high-frequency detail components H 1 、H 2 and H 3 ; H 1 has a scale of 1, H 2 has a scale of 2, H 3 has a scale of 3;

[0038] For the high-frequency detail component H i of the i-th layer, in the 3×3 neighborhood centered on the pixel point (x, y), calculate the local standard deviation:

[0039]

[0040] By performing summation, averaging, and squaring operations on the high-frequency detail component H i within the 3×3 neighborhood to obtain the local standard deviation and the local mean. In the formula, s i (x, y) represents the local standard deviation of the 3×3 neighborhood centered on (x, y), m i (x, y) represents the local average of the 3×3 neighborhood centered on (x, y), m represents the spatial index in the horizontal direction, and n represents the spatial index in the vertical direction;

[0041] Calculate the adaptive threshold T i (x, y) = a i 's i (x, y) + b i 'm i (x, y), where a i and b i are adjustment parameters related to the scale of the high-frequency detail component. At scale 1, a 1 = 1.5, b 1 = 0.5; at scale 2, a 2 = 1.2, b 2 = 0.3; at scale 3, a 3 = 1.0, b 3 = 0.2. The adaptive threshold is obtained by multiplying the local standard deviation and the local mean by the corresponding adjustment parameters and then adding them;

[0042] For the high-frequency detail component H i of each scale, compare the amplitude of the wavelet coefficient |Hi (x, y) and the corresponding adaptive threshold T i (x, y): If |H i (x, y)| > T i Then mark the pixel as a possible edge point with a value of 1; otherwise, mark it as a non-edge point with a value of 0, and thus obtain the preliminary edge detection result E at each scale 1 、E 2 、E 3 ;

[0043] By multiplying the preliminary edge detection results at different scales by the corresponding weights and then summing them up, calculate the edge intensity E(x, y) of each pixel point (x, y). The formula is: E(x, y) = w 1 ′E 1 (x, y) + w 2 ′E 2 (x, y) + w 3 ′E 3 (x, y), and obtain the fused edge intensity image

[0044] Furthermore, the specific method of using the Laplacian operator with increased horizontal direction weight to perform edge detection on the denoised image is as follows:

[0045] The Laplacian operator with increased horizontal direction weight is:

[0046]

[0047] For a 3x3 pixel region Use the Laplacian operator with increased horizontal direction weight to perform convolution operation on the central pixel I 22 , and the central pixel I 22 The new value after being processed by this operator is:

[0048]

[0049] Replace the obtained with the original central pixel I 22 , if then it is considered an edge point and the pixel is set to 255, if then it is considered not an edge point and the pixel is set to 0. Calculate each point in turn to obtain the edge binary image

[0050] Furthermore, the specific method of calculating the SSIM value of the first edge image and the second edge image is as follows:

[0051] The calculation formula of the structural similarity index SSIM is as follows:

[0052]

[0053] In the formula, e and f respectively represent window regions of size m′n at the same position in two edge images after being detected by the adaptive multi-scale wavelet edge detection algorithm and the Laplacian operator with increased horizontal direction weight; SSIM(e,f) represents the structural similarity index between region e and region f, m e represents the mean value of region e, m f represents the mean value of region f, C 1 、C 2 are constants, is the variance of region e, is the variance of region f, s ef is the covariance of regions e and f.

[0054] Furthermore, the dynamic adjustment of the fusion weight based on the SSIM value and the weighted fusion of the first edge image and the second edge image are specifically as follows:

[0055] When SSIM is greater than the set threshold, the pixels of the first edge image and the second edge image are respectively multiplied by 0.7 and then added, and the result is limited between 0 and 255; when SSIM is less than or equal to the set threshold, the pixels of the first edge image and the second edge image are respectively multiplied by 0.3 and then added.

[0056] The present invention also proposes a device for detecting and adjusting the expansion of a tubular belt conveyor, including:

[0057] An image acquisition unit for taking pictures of the tubular belt conveyor;

[0058] An industrial control computer, including: a distortion correction module, a denoising module, an edge detection module, and a judgment module;

[0059] The distortion correction module is used to perform perspective transformation on the picture of the tubular belt conveyor and convert the viewing angle of the picture of the tubular belt conveyor to a top view angle;

[0060] The denoising module converts the colored picture of the tubular belt conveyor into a grayscale image and performs denoising processing on the grayscale image using bilateral filtering;

[0061] The edge detection module is used to perform edge detection on the denoised image through an adaptive multi-scale wavelet edge detection algorithm and binarize the result to obtain a first edge picture; perform edge detection on the denoised image using a Laplacian operator with increased horizontal direction weights to obtain a second edge picture; calculate the SSIM value of the first edge picture and the second edge picture, dynamically adjust the fusion weights based on the SSIM value, and perform weighted fusion on the first edge picture and the second edge picture; use the LSD algorithm to fit two best edge lines in the fused edge picture as the pipe belt edge lines.

[0062] The judgment module is used to calculate the maximum distance between the pipe belt edge lines. If the maximum distance is greater than or equal to the threshold, it is judged as tube expansion, and a tube expansion signal is sent to the motor.

[0063] The motor is used to drive the electric push rod to expand and contract after receiving the tube expansion signal.

[0064] The electric push rod is used to drive the idler to descend to increase the pipe diameter.

[0065] The beneficial effects of the present invention are as follows:

[0066] 1. The present invention provides an improved wavelet edge detection method, which improves the accuracy of the pipe belt machine edge detection by combining a unidirectional edge detection operator.

[0067] 2. By combining perspective transformation with the LSD straight line fitting algorithm, the actual positions of the pipe belt and its edge are detected in real time without loss of accuracy.

[0068] 3. Compared with the existing tube expansion detection device, it avoids the wear caused by contact with the pipe belt, and provides a more reliable solution for the safety management of the pipe belt machine.

[0069] 4. The tube expansion problem of the pipe belt machine is solved in real time through the pipe diameter adjustment device, reducing the labor cost and avoiding further economic losses caused by tube expansion. Description of the Drawings

[0070] Figure 1 It is the overall flowchart of the tube expansion detection and adjustment method for the tubular belt conveyor proposed by the present invention.

[0071] Figure 2 It is the architecture diagram of the tube expansion detection and adjustment device for the tubular belt conveyor.

[0072] Figure 3 It is the flowchart of detecting the pipe belt edge by combining multi-scale adaptive wavelet edge detection with a unidirectional edge detection operator.

[0073] Figure 4 It is the effect diagram of the detected pipe belt edge line. Detailed Implementation Modes

[0074] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present application without creative efforts shall fall within the protection scope of the present application.

[0075] Embodiment 1

[0076] The present invention provides a method for detecting and adjusting the expansion of a tubular belt conveyor. The flowchart of this method is as Figure 1 shown and includes the following steps:

[0077] S1. Take pictures of the tubular belt conveyor;

[0078] S2. Perform perspective transformation on the pictures of the tubular belt conveyor. The perspective transformation is based on the principle of projective geometry to convert the viewing angle of the pictures of the tubular belt conveyor to the top-down viewing angle. Specifically:

[0079] Select four groups of idler vertices on both sides of the pipe belt as 4 points (x i , y i ) in the source image and 4 points (x i ', y i ') in the target image. The perspective transformation matrix H is obtained by solving the following linear equations:

[0080]

[0081] In the formula, h ij is the element in the i-th row and j-th column of the perspective transformation matrix,

[0082] The perspective transformation matrix H has 8 degrees of freedom because the matrix element h 33 is usually set to 1 to ensure the normalization of homogeneous coordinates.

[0083] Perform perspective transformation on the point (x, y) on the pictures of the tubular belt conveyor through the following transformation formula:

[0084]

[0085] In the formula, (x’, y’) is the position of the new point after mapping.

[0086] S3. Convert the colored pictures of the tubular belt conveyor into grayscale images, and perform denoising processing on the grayscale images using bilateral filtering. Specifically:

[0087] For each pixel p in the grayscale image, its grayscale value I after bilateral filteringbf (p) is calculated by the following formula:

[0088]

[0089] where: S is the neighborhood centered on pixel p, usually a square region, is the Gaussian kernel function in the spatial domain, used to measure the distance weight between pixels p and q in the spatial position, is the Gaussian kernel function in the gray value range, used to measure the difference weight between pixels p and q in the gray value. I(p) represents the original gray value of pixel q in the neighborhood S, and I(q) is the original gray value of pixel q in the neighborhood S, where q is any pixel point in the neighborhood S where p is located and not equal to p, is the normalization factor to ensure that the filtered pixel value is within a reasonable range.

[0090] The expression of the Gaussian kernel function in the spatial domain is: where d = ||p - q|| is the Euclidean distance between pixels p and q, with pixel coordinates (x p , y p ) and (x q , y q ) representing the coordinates of pixels p and q, then s d is the standard deviation of the Gaussian kernel in the spatial domain, which controls the degree of influence of the spatial distance on the weight. The larger s d , the wider the influence range of the pixels in the spatial neighborhood on the central pixel.

[0091] The expression of the Gaussian kernel function in the gray value range is: where DI = |I(p) - I(q)| is the difference in gray values between pixels p and q, and s r is the standard deviation of the Gaussian kernel in the gray value range, which controls the degree of influence of the gray value difference on the weight. The smaller s r , only pixels with similar gray values will have a greater influence on the central pixel, thus better retaining edge information.

[0092] Calculate for each pixel q in the neighborhood S, then sum and divide by the normalization factor W p to obtain I bf (p). Calculate each pixel in the image according to this method to complete bilateral filtering denoising.

[0093] S4. Perform edge detection on the denoised image through the adaptive multi-scale wavelet edge detection algorithm and binarize the result to obtain the first edge picture; Performing edge detection on the denoised image through the adaptive multi-scale wavelet edge detection algorithm specifically includes:

[0094] Select Symlets wavelets for three-level wavelet decomposition. In each level of decomposition, convolve the image with the low-pass filter h and the high-pass filter g. For the horizontal direction:

[0095]

[0096] For the vertical direction:

[0097]

[0098] where I i-1 is the image of the previous level, s represents the spatial index of I i-1 in the horizontal direction, t represents the spatial index of I i-1 in the vertical direction, m represents the spatial index of I i after convolution in the horizontal direction of the component obtained, n represents the spatial index of I i after convolution in the vertical direction of the component obtained, LL i is the low-frequency approximation component of the i-th level, LH i represents the value of the component combining low frequency in the horizontal direction and high frequency in the vertical direction in the i-th level wavelet decomposition, HL i is the value of the component combining high frequency in the horizontal direction and low frequency in the vertical direction in the i-th level wavelet decomposition, HH i represents the value of the component with high frequency in both the horizontal and vertical directions in the i-th level wavelet decomposition, N is the size of the image; after three-level decomposition, high-frequency detail components H 1 、H 2 and H 3 ; H 1 has a scale of 1, H 2 has a scale of 2, H 3 has a scale of 3;

[0099] For the high-frequency detail component H i of the i-th level, calculate the local standard deviation for the 3×3 neighborhood centered on the pixel point (x, y):

[0100]

[0101] By performing summation, averaging, and squaring operations on the high-frequency detail component H i within the 3×3 neighborhood, obtain the local standard deviation and the local mean. In the formula, s i (x, y) represents the local standard deviation of the 3×3 neighborhood centered on (x, y), m i (x, y) represents the local average of the 3×3 neighborhood centered on (x, y), m represents the spatial index in the horizontal direction, and n represents the spatial index in the vertical direction;

[0102] Calculate the adaptive threshold \(T\) through \(T(x,y)=a\times s(x,y)+b\), where \(a\) and \(b\) are adjustment parameters related to the scale of the high - frequency detail component. At scale 1, \(a = 1.5\) and \(b = 0.5\); at scale 2, \(a = 1.2\) and \(b = 0.3\); at scale 3, \(a = 1.0\) and \(b = 0.2\). The adaptive threshold is obtained by multiplying the local standard deviation and the local mean by the corresponding adjustment parameters and then adding them together. i (x,y)=a i ′s i (x,y)+b i ′m i (x,y) calculate the adaptive threshold \(T\), where \(a\) i and \(b\) i are adjustment parameters related to the scale of the high - frequency detail component. At scale 1, \(a\) 1 =1.5, \(b\) 1 =0.5; at scale 2, \(a\) 2 =1.2, \(b\) 2 =0.3; at scale 3, \(a\) 3 =1.0, \(b\) 3 =0.2. The adaptive threshold is obtained by multiplying the local standard deviation and the local mean by the corresponding adjustment parameters and then adding them together;

[0103] For the high - frequency detail component \(H\) of each scale i , compare the magnitude of the wavelet coefficient \(|H(x,y)|\) of each pixel with the corresponding adaptive threshold \(T_i(x,y)\): i (x,y)| with the corresponding adaptive threshold \(T_i(x,y)\): If \(|H(x,y)|>T(x,y)\), then mark this pixel as a possible edge point with a value of 1; otherwise, mark it as a non - edge point with a value of 0. In this way, the preliminary edge detection results \(E_1\), \(E_2\), \(E_3\) at each scale are obtained; i (x,y)|>T i (x,y) then mark this pixel as a possible edge point, set the value to 1; otherwise, mark it as a non - edge point, set the value to 0, so as to obtain the preliminary edge detection results \(E\) 1 、\(E\) 2 、\(E\) 3 ;

[0104] Calculate the edge intensity \(E(x,y)\) of each pixel \((x,y)\) by multiplying the preliminary edge detection results at different scales by the corresponding weights and then adding them together. The formula is: \(E(x,y)=w_1\times E_1(x,y)+w_2\times E_2(x,y)+w_3\times E_3(x,y)\), and the fused edge intensity image is obtained. 1 ′E 1 (x,y)+w 2 ′E 2 (x,y)+w 3 ′E 3 (x,y), obtain the fused edge intensity image.

[0105] S5. Use the Laplacian operator with increased horizontal - direction weight to perform edge detection on the denoised image to obtain the second edge picture; specifically:

[0106] We use \(I(x,y)\) to represent the pixel value of the image at the coordinate \((x,y)\). Assuming \(D_x = D_y = 1\) (pixel spacing), the discrete approximation of the second - order partial derivative in the \(x\) - direction is: For the traditional edge - detection operator, in the two - dimensional case, it is It simultaneously considers the second-order variations in both the horizontal and vertical directions. To detect only the horizontal direction, we retain only the second-order partial derivative part in the x direction, and to highlight the edges in the horizontal direction, we perform a weighting process. Design a 3x3 Laplacian operator template According to the discrete form of the second-order partial derivative in the x direction mentioned above, we hope that this template can simulate the effect of (I(i,j+1)-2I(i,j)+I(i,j-1)) during calculation, while highlighting the edge changes in the horizontal direction through weighting.

[0107] The Laplacian operator with increased horizontal direction weight designed in this embodiment is:

[0108]

[0109] For a 3x3 pixel region Use the Laplacian operator with increased horizontal direction weight to perform a convolution operation on the central pixel I 22 The new value of the central pixel I 22 after being processed by this operator is:

[0110]

[0111] Replace the obtained with the original central pixel I 22 , if then it is considered an edge point and the pixel is set to 255, if then it is considered not an edge point and the pixel is set to 0. Calculate each point in turn to obtain the edge binary image.

[0112] S6. Calculate the SSIM value of the first edge image and the second edge image. The calculation formula of the structural similarity index SSIM is as follows:

[0113]

[0114] In the formula, e and f respectively represent the window regions of size m′n at the same position in the two edge images after being detected by the adaptive multi-scale wavelet edge detection algorithm and the Laplacian operator with increased horizontal direction weight; SSIM(e,f) represents the structural similarity index of region e and region f, m e represents the mean value of region e, m f represents the mean value of region f, C 1 、C 2 are constants, is the variance of region e, is the variance of region f, s efIt is the covariance of regions e and f. The fusion weights are dynamically adjusted based on the SSIM value, and the first edge image and the second edge image are weighted and fused. When the SSIM is greater than the set threshold, it indicates that the structural similarity of the detection results of the two methods is relatively large. The pixels of the first edge image and the second edge image are multiplied by 0.7 respectively and then added together, and the result is limited between 0 and 255. When the SSIM is less than or equal to the set threshold, it indicates that the detection results of the two methods are quite different. The pixels of the first edge image and the second edge image are multiplied by 0.3 respectively and then added together.

[0115] S7. Fit two best edge lines in the fused edge image using the LSD algorithm as the edge lines of the pipe belt. Calculate the level-line and its angle for each pixel, and convert the image into a level-line field. According to the level-line angle, the pixels within the tolerance are classified into line-support regions as line segment candidates. Then, screen the candidates, construct the minimum circumscribed rectangle with its principal inertia axis, and check the level-line angles of the pixels within the rectangle. The pixels with an angle difference within the tolerance are homogeneous points. Count the number of pixels and homogeneous points within the rectangle, and use the contrario approach and the Helmholtz principle to detect the line. Introduce a contrario model (a perfect noise image, with pixel values independent and uniformly distributed in [0, 2π]), count the pixels and aligned points within the rectangles of the target and noise images, calculate the probability that the aligned points in the target image are fewer than those in the noise image and multiply by the proportionality coefficient to obtain the NFA value, and judge whether it is a line based on this.

[0116] S8. Calculate the maximum distance between the edge lines of the pipe belt. If the maximum distance is greater than or equal to the threshold, it is judged as tube expansion. When the pipeline diameter expands by about 3% - 5%, it may be necessary to pay attention to whether tube expansion occurs. Traverse the Cartesian coordinates of the points on the two edge segments of the pipe belt respectively. At the same x value, calculate |y 1 -y 2 | to obtain the Euclidean distance between the two points. Assume that when the tube expansion operates normally, the distance between the two edges is P. If then it is judged as tube expansion.

[0117] S9. Increase the pipe diameter after receiving the tube expansion signal.

[0118] Embodiment 2

[0119] The present invention provides a tube expansion detection and adjustment device corresponding to the method of Embodiment 1, including:

[0120] An image acquisition unit for taking pictures of the tubular belt conveyor; in this embodiment, the image acquisition unit is a camera, which is connected to the computer through relevant interfaces, and a video acquisition program based on OpenCV is written on the computer to realize the real-time acquisition of video.

[0121] An industrial control computer, including: a distortion correction module, a denoising module, an edge detection module, and a judgment module;

[0122] The distortion correction module is used to perform perspective transformation on the pictures of the tubular belt conveyor and convert the perspective of the pictures of the tubular belt conveyor to a top view perspective;

[0123] The denoising module converts the colored pictures of the tubular belt conveyor into grayscale images and performs denoising processing on the grayscale images using bilateral filtering;

[0124] The edge detection module is used to perform edge detection on the denoised image through an adaptive multi-scale wavelet edge detection algorithm and binarize the result to obtain a first edge picture; perform edge detection on the denoised image using a Laplacian operator with increased horizontal direction weights to obtain a second edge picture; calculate the SSIM value of the first edge picture and the second edge picture, dynamically adjust the fusion weight based on the SSIM value, and perform weighted fusion on the first edge picture and the second edge picture; use the LSD algorithm to fit two best edge straight lines in the fused edge picture as the pipe belt edge straight lines;

[0125] The judgment module is used to calculate the maximum distance between the pipe belt edge straight lines. If the maximum distance is greater than or equal to the threshold, it is judged as tube expansion, and a tube expansion signal is sent to the motor;

[0126] The motor is used to drive the electric push rod to expand and contract after receiving the tube expansion signal;

[0127] The electric push rod is used to drive the idler to descend to increase the pipe diameter.

[0128] The implementation methods of each module and module functions in the system are exactly the same as the steps of the method in Embodiment 1, so they will not be elaborated here.

[0129] The structure of the tube expansion detection and adjustment device for the tubular belt conveyor is as Figure 2 shown. The tube expansion detection and adjustment device equipped with the camera 1 and the industrial control computer 2 is installed between two groups of idlers of the pipe belt machine. When tube expansion is detected, the industrial control computer 2 sends a tube expansion signal to the motor 3, the motor 3 forms a circuit, and the electric push rod 4 drives the idler 5 to slowly descend, increasing the pipe diameter, so that the same flow of materials has more space to flow, alleviating the tube expansion phenomenon.

[0130] In the embodiments disclosed in the present application, the computer storage medium may be a tangible medium that can contain or store a program for use by or in connection with an instruction execution system, apparatus, or device. The computer storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of the computer storage medium would include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0131] Those of ordinary skill in the art will appreciate that the units and algorithm steps of the examples described in connection with the embodiments disclosed in the present application can be implemented in electronic hardware or in a combination of computer software and electronic hardware. Whether these functions are executed in hardware or software depends on the specific application and design constraints of the technical solution. Skilled artisans may use different methods for each specific application to implement the described functions, but such implementation should not be considered to exceed the scope of the present application.

[0132] The above are only the preferred embodiments of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions falling within the idea of the present invention belong to the protection scope of the present invention. It should be noted that for those of ordinary skill in the art in the technical field, several improvements and refinements made without departing from the principle of the present invention should be regarded as within the protection scope of the present invention.

Claims

1. A method for detecting and adjusting expansion of a tubular belt conveyor, characterized in that: The following steps are involved: Take pictures of tubular belt conveyors; Perform perspective transformation on the tubular belt conveyor picture, and convert the perspective of the tubular belt conveyor picture to a top-down perspective; The colored tubular belt conveyor picture is converted into a grayscale image, and the grayscale image is denoised using bilateral filtering; Perform edge detection on the denoised image using an adaptive multi-scale wavelet edge detection algorithm and binarize the result to obtain a first edge image; The denoised image is subjected to edge detection using a Laplacian operator with increased horizontal weight, thereby obtaining a second edge image. Calculate the SSIM value of the first edge image and the second edge image, dynamically adjust the fusion weight based on the SSIM value, and perform weighted fusion on the first edge image and the second edge image; Use the LSD algorithm to fit the two best edge lines in the fused edge image as the tube edge lines; Calculate the maximum distance between the straight lines at the edge of the tube strip. If the maximum distance is greater than or equal to the threshold, it is judged as tube expansion. Increase the tube diameter after receiving the tube expansion signal.

2. The method for detecting and adjusting the expansion of a tubular belt conveyor according to claim 1, characterized in that: The perspective transformation of the tubular belt conveyor picture is specifically as follows: Select four groups of roller vertices on both sides of the pipe belt as the four points in the source image (x i ,y i ) and the four points (x i ',y i '), the perspective transformation matrix H is obtained by solving the following linear equations: In the formula, h ij is the element in the i-th row and j-th column of the perspective transformation matrix, The perspective transformation of the point (x, y) on the tubular belt conveyor image is performed using the following transformation formula: Where (x', y') is the position of the new point after mapping.

3. The method for detecting and adjusting the expansion of a tubular belt conveyor according to claim 1, characterized in that: The denoising process of the grayscale image using bilateral filtering is specifically as follows: For each pixel p in the grayscale image, its grayscale value I after bilateral filtering bf (p) is calculated by the following formula: Among them: S is the neighborhood centered on pixel p, It is a Gaussian kernel function in the spatial domain, which is used to measure the distance weight of pixels p and q in spatial position. is a Gaussian kernel function in the grayscale value domain, which is used to measure the difference weight between pixels p and q in grayscale value. I(p) represents the original grayscale value of pixel q in the neighborhood S, and I(q) represents the original grayscale value of pixel q in the neighborhood S, where q is any pixel point in the neighborhood S where p is located that is not equal to p. is the normalization factor; The expression of the Gaussian kernel function in the spatial domain is: where d = ||pq|| is the Euclidean distance between pixels p and q, s d is the standard deviation of the Gaussian kernel in the spatial domain; The expression of the gray value domain Gaussian kernel function is: Where DI = |I(p)-I(q)| is the difference between the grayscale values ​​of pixels p and q, s r is the standard deviation of the Gaussian kernel in the grayscale value domain; For each pixel q in the neighborhood S, calculate The values ​​are then summed and divided by the normalization factor W p Get I bf (p), calculate each pixel in the image according to this method to complete bilateral filtering denoising.

4. The method for detecting and adjusting the expansion of a tubular belt conveyor according to claim 1, characterized in that: The edge detection of the denoised image by using the adaptive multi-scale wavelet edge detection algorithm is specifically as follows: The Symlets wavelet is selected for three-layer wavelet decomposition. In each layer of decomposition, the image is convolved by a low-pass filter h and a high-pass filter g. For the horizontal direction: For vertical direction: Among them, I i-1 is the image of the previous layer, s represents I i-1 Spatial index in the horizontal direction, t represents I i-1 Spatial index in the vertical direction, m represents I i The spatial index of the component obtained after the convolution operation in the horizontal direction, n represents I i The spatial index of the component obtained after the convolution operation in the vertical direction, LL i is the low-frequency approximate component of the i-th layer, LH i Represents the value of the component of the combination of horizontal low frequency and vertical high frequency in the i-th layer wavelet decomposition, HL i is the value of the component of the combination of horizontal high frequency and vertical low frequency in the i-th layer wavelet decomposition, HH i It represents the value of the high-frequency component in both horizontal and vertical directions in the i-th layer of wavelet decomposition, and N is the size of the image. After three layers of decomposition, each layer obtains high-frequency detail components H1, H2 and H3 of different scales. The scale of H1 is 1, the scale of H2 is 2, and the scale of H3 is 3. For the high-frequency detail component H of the i-th layer i , calculate the local standard deviation in a 3′3 neighborhood centered on the pixel point (x, y): By calculating the high-frequency detail component H in the 3′3 neighborhood i The local standard deviation and local mean are obtained by summing, averaging and squaring, where s i (x,y) represents the local standard deviation of the 3′3 neighborhood centered at (x,y), m i (x, y) represents the local average of the 3′3 neighborhood centered at (x, y), m represents the spatial index in the horizontal direction, and n represents the spatial index in the vertical direction; By T i (x,y)=a i 's i (x,y)+b i ′m i (x, y) calculates the adaptive threshold T, where a i and b i It is an adjustment parameter related to the scale of the high-frequency detail component. At scale 1, a1=1.5, b1=0.5; at scale 2, a2=1.2, b2=0.3; at scale 3, a3=1.0, b3=0.

2. The adaptive threshold is obtained by multiplying the local standard deviation and the local mean with the corresponding adjustment parameters and then adding them together; For each scale high-frequency detail component H i , compare the amplitude of the wavelet coefficient of each pixel |H i (x,y)| and the corresponding adaptive threshold T i (x,y): If |H i (x,y)|>T i (x, y), the pixel is marked as a possible edge point and the value is set to 1; otherwise, it is marked as a non-edge point and the value is set to 0. In this way, the preliminary edge detection results E1, E2, and E3 are obtained at each scale; The edge intensity E(x,y) of each pixel (x,y) is calculated by multiplying the preliminary edge detection results at different scales by the corresponding weights and then adding them. The formula is: E(x,y)=w1′E1(x,y)+w2′E2(x,y)+w3′E3(x,y), and the fused edge intensity image is obtained.

5. The method for detecting and adjusting the expansion of a tubular belt conveyor according to claim 1, characterized in that: The edge detection of the denoised image using the Laplacian operator with increased horizontal weight is specifically as follows: The Laplacian operator with increased horizontal weight is: For a 3x3 pixel area The Laplacian operator with increased horizontal weight is used to calculate the center pixel I 22 Perform convolution operation, the center pixel I 22 The new value after being processed by the operator for: Will get Replace the original center pixel I 22 ,like It is considered to be an edge point and the pixel is set to 255. If it is not an edge point, the pixel is set to 0, and each point is calculated in turn to obtain an edge binary map.

6. The method for detecting and adjusting the expansion of a tubular belt conveyor according to claim 1, characterized in that: The calculation of the SSIM values ​​of the first edge picture and the second edge picture is specifically as follows: The calculation formula of the structural similarity index SSIM is as follows: Where e and f represent the window area of ​​size m′n at the same position of the two edge images after detection by the adaptive multi-scale wavelet edge detection algorithm and the Laplace operator with increased horizontal weight; SSIM(e,f) represents the structural similarity index of area e and area f, m e represents the mean value of region e, m f represents the mean value of region f, C1 and C2 are constants, is the variance of region e, is the variance of region f, s ef is the covariance of regions e and f.

7. The method for detecting and adjusting the expansion of a tubular belt conveyor according to claim 1, characterized in that: The method of dynamically adjusting the fusion weight based on the SSIM value and performing weighted fusion on the first edge picture and the second edge picture is specifically as follows: When SSIM is greater than the set threshold, the pixels of the first edge image and the second edge image are multiplied by 0.7 and added together, and the result is limited to between 0 and 255; when SSIM is less than or equal to the set threshold, the pixels of the first edge image and the second edge image are multiplied by 0.3 and added together.

8. A tubular belt conveyor expansion detection and adjustment device, characterized in that: include: An image acquisition unit, used for taking pictures of the tubular belt conveyor; Industrial computer, including: distortion correction module, denoising module, edge detection module and judgment module; The distortion correction module is used to perform perspective transformation on the tubular belt conveyor picture, and convert the viewing angle of the tubular belt conveyor picture to a top-down viewing angle; The denoising module converts the color tubular belt conveyor picture into a grayscale image and performs denoising on the grayscale image using bilateral filtering; The edge detection module is used to perform edge detection on the denoised image by using an adaptive multi-scale wavelet edge detection algorithm and binarize the result to obtain a first edge image; perform edge detection on the denoised image by using a Laplace operator with a horizontal weight added to obtain a second edge image; calculate the SSIM value of the first edge image and the second edge image, dynamically adjust the fusion weight based on the SSIM value, and perform weighted fusion on the first edge image and the second edge image; use the LSD algorithm to fit two optimal edge lines in the fused edge image as the pipe band edge lines; The judgment module is used to calculate the maximum distance between the straight lines at the edges of the tube belt, and if the maximum distance is greater than or equal to a threshold, it is judged as tube expansion, and a tube expansion signal is sent to the motor; The motor is used to drive the electric push rod to extend and retract after receiving the tube expansion signal; The electric push rod is used to drive the roller to descend to increase the pipe diameter.

Images