A method and system for detecting belt tear conditions based on image recognition

Abstract

A method and system for detecting belt tearing condition based on image recognition, wherein the method comprises the following steps: acquiring a running image of the belt, and setting a ROI region containing a laser in the running image; cropping the original image of the ROI region of each frame in the running image; fitting the ROI region of each frame of the original image to obtain the upper boundary and the lower boundary of the laser; obtaining a suspicious tearing point group through the height difference between the upper boundary and the lower boundary; obtaining the curvature of the suspicious tearing point group and its adjacent pixel points, counting the number of times the tearing point appears in the running image within a width range, and obtaining the tearing length of the belt through the running speed of the belt and the number of frames of the tearing point appearing. The present application intelligently detects the tearing condition of the belt through image recognition, compared with the prior art of determining the tearing degree of the belt by artificial means, can effectively improve the accuracy of measuring the belt tearing, and timely warns of the belt breakage.

Description

Technical Field

[0001] This invention relates to the field of image recognition technology, and in particular to a method and system for detecting belt tears based on image recognition. Background Technology

[0002] Belt conveyors are important transportation equipment for transporting materials in many industries, including ports, metallurgy, mining, chemical, petroleum, power plants, and building materials. The main materials of the conveyor belts are rubber and fiber core belts, ordinary canvas core belts, or steel rope core belts, etc. However, conveyor belts are prone to tearing during continuous operation.

[0003] When a conveyor belt breaks during transport, the falling material may cause spillage, damage to reducers and motors, and other equipment. Due to their cost, conveyor belts cannot be replaced at the slightest tear. Furthermore, the breakage process is real-time, and timed inspections cannot effectively predict when replacement is needed; only real-time monitoring can accurately detect belt tears. However, conveyor belts located above the conveyor are usually loaded with goods, hindering worker observation and resulting in low accuracy in determining the extent of tearing. Because of their position, workers cannot easily inspect conveyor belts from below. Therefore, a method for automatically identifying belt tears is urgently needed. Summary of the Invention

[0004] To address the aforementioned shortcomings, the present invention aims to propose a method and system for detecting belt tears based on image recognition, thereby achieving automatic identification of belt tears and calculation of tear length, and solving the problem that it is difficult for humans to observe belt tears during operation.

[0005] To achieve this objective, the present invention employs the following technical solution: a method for detecting belt tearing based on image recognition, comprising the following steps:

[0006] Step S1: Use a laser to scan the belt to obtain a running image of the belt, and select a ROI region containing the laser in the running image;

[0007] Step S2: Crop the original image of the ROI region for each frame in the running image;

[0008] Step S3: Fit the ROI region of each frame of the original image to obtain the upper and lower boundaries of the laser.

[0009] Step S4: Obtain the group of suspected tear points by using the height difference between the upper and lower boundaries;

[0010] Step S5: Obtain the curvature of the suspected tear point group and its adjacent pixels, and determine whether the curvature of the suspected tear point group meets the curvature threshold. If it does, then the point is taken as the tear point.

[0011] Step S6: Count the number of times the tear point appears in the running image within the width range, and obtain the tear length of the belt by the running speed of the belt and the number of frames in which the tear point appears.

[0012] Preferably, the specific steps of step S3 are as follows:

[0013] Step S1: Use the cv2.medianBlur function to perform median filtering on the original image;

[0014] Step S2: Binarize the filtered original image to obtain the approximate outline of the laser in the original image;

[0015] Step A3: Perform region growing or connected region processing on the general outline of the laser to obtain the largest laser area;

[0016] Step S4: Perform edge detection on the largest laser area to obtain the laser contour;

[0017] Step S5: Use a curve fitting algorithm to obtain the upper and lower boundaries of the laser profile.

[0018] Preferably, the specific steps of step S4 are as follows:

[0019] Step S41: Traverse the upper and lower boundaries, and store the pixel coordinates of the upper and lower boundaries in two coordinate tables respectively.

[0020] Step S42: Treat two pixels with the same horizontal coordinate in the two coordinate tables as a group of judgment points;

[0021] Step S43: Obtain the difference in the ordinates of two pixels in the judgment point group. If the difference in the ordinates of two pixels is greater than the difference threshold, then set the judgment point group as a suspected tear point group.

[0022] Preferably, the specific steps of step S5 are as follows:

[0023] Step S51: Obtain the adjacent pixels on the left and right sides of two pixels in the suspected tear point group respectively, and obtain the distance t from the two adjacent pixels to the pixels in the suspected tear point group respectively. a and distance t b ;

[0024] Step S52: Construct a quadratic function model: y = a0 + a1x + a2x 2 ;

[0025] Using distance t a and distance t b The first matrix is ​​constructed as follows:

[0026]

[0027] Step S53: Calculate the specific values ​​of the coefficient vectors a0, a1, and a2 using the first matrix and a quadratic function model;

[0028] Step S54: Obtain the curvature of the pixels in the suspicious point group using coefficient vectors a1 and a2, where the formula for obtaining the curvature is as follows: Where b1 = a1 + 2a2x2, b2 = 2a2, and x2 is the x-coordinate of one of the adjacent pixels;

[0029] Step S55: Determine whether the curvature of any pixel in the suspicious point group is greater than the curvature threshold. If it is greater, mark the suspicious point group as a tear point.

[0030] Preferably, the specific steps of step S6 are as follows:

[0031] Once a tear point first appears in the current frame, mark the tear point and set the count of the marked tear point to 1.

[0032] Determine if a tear point appears in the next frame. If a tear point appears, obtain the difference between the x-coordinate of the tear point in this frame and the x-coordinates of all marked tear points in the previous frame. Determine if the difference of the x-coordinates is greater than the width range. If the difference of all x-coordinates is greater than the width range, mark the tear point and set the occurrence count of the marked tear point to 1. If there is a difference of x-coordinates that is less than the width range, increase the occurrence count of the marked tear points within the width range. If no tear point appears, determine if the occurrence count of the tear point has been updated. If the occurrence count has not been updated, calculate the length of the tear.

[0033] The formula for calculating the tear length is as follows: S = n * v / e, where n is the number of times the tear point occurs, v is the speed of the belt, and e is the number of frames.

[0034] A system for detecting belt tears based on image recognition, using the aforementioned method for detecting belt tears based on image recognition, includes: a selection module, a cropping module, a fitting module, a judgment module, a determination module, and a length calculation module;

[0035] The selection module is used to scan the belt with a laser to obtain a running image of the belt, and to select a ROI region containing the laser in the running image;

[0036] The cropping module is used to crop out the original image of the ROI region of each frame in the running image;

[0037] The fitting module is used to fit the ROI region of each frame of the original image to obtain the upper and lower boundaries of the laser.

[0038] The judgment module is used to obtain a group of suspected tear points by using the height difference between the upper and lower boundaries;

[0039] The determining module is used to obtain the curvature of the suspected tear point group and its adjacent pixels, and to determine whether the curvature of the suspected tear point group meets the curvature threshold. If it does, the point is taken as the tear point.

[0040] The length calculation module is used to count the number of times the tear point appears in the running image within the width range, and to obtain the tear length of the belt by the running speed of the belt and the number of frames in which the tear point appears.

[0041] Preferably, the fitting module includes a first processing submodule, a second processing submodule, a third processing submodule, a fourth processing submodule, and a fifth processing submodule;

[0042] The first processing submodule is used to perform median filtering on the original image using the cv2.medianBlur function;

[0043] The second processing submodule is used to binarize the filtered original image to obtain the approximate outline of the laser in the original image;

[0044] The three processing submodules are used to perform region growth or connected region processing on the general outline of the laser to obtain the largest laser region.

[0045] The fourth processing submodule is used to perform edge detection on the largest laser area to obtain the laser contour.

[0046] The fifth processing submodule is used to apply a curve fitting algorithm to the laser profile to obtain the upper and lower boundaries of the laser.

[0047] Preferably, the judgment module includes a traversal submodule, a grouping submodule, and a difference calculation submodule;

[0048] The traversal submodule is used to traverse the upper and lower boundaries, and store the pixel coordinates of the upper and lower boundaries after traversal in two coordinate tables respectively.

[0049] The grouping submodule is used to treat two pixels with the same horizontal coordinate in two coordinate tables as a group of judgment points.

[0050] The difference calculation submodule is used to obtain the difference in the ordinate of two pixels in the judgment point group. If the difference in the ordinate of two pixels is greater than the difference threshold, the judgment point group is set as a suspected tear point group.

[0051] Preferably, the determining module includes a first calculation submodule, a second calculation submodule, a third calculation submodule, a fourth calculation submodule, and a fifth calculation submodule;

[0052] The first calculation submodule is used to obtain the adjacent pixels on the left and right sides of two pixels in the suspected tear point group, and to obtain the distance t from the two adjacent pixels to the pixels in the suspected tear point group. a and distance t b ;

[0053] The second calculation submodule is used to construct a quadratic function model: y = a0 + a1x + a2x 2 ;

[0054] Using distance t a and distance t b Construct the first matrix;

[0055] The third calculation submodule is used to calculate the specific values ​​of the coefficient vectors a0, a1 and a2 by using the first matrix and a quadratic function model.

[0056] The fourth calculation submodule is used to obtain the curvature of the pixels in the suspicious point group through coefficient vectors a1 and a2;

[0057] The fifth calculation submodule is used to determine whether the curvature of any pixel in the suspicious point group is greater than the curvature threshold. If it is greater, the suspicious point group is marked as a tear point.

[0058] Preferably, the length calculation module includes a first analysis submodule and a second analysis submodule;

[0059] The method is to mark the tear point after it first appears in the current frame, and set the number of times the marked tear point appears to be 1.

[0060] The second analysis submodule is used to determine whether a tear point appears in the next frame. If a tear point appears, the difference between the x-coordinate of the tear point in this frame and the x-coordinate of all marked tear points in the previous frame is obtained one by one. It is then determined whether the difference of the x-coordinate is greater than the width range. If the difference of all the x-coordinates is greater than the width range, the tear point is marked and the occurrence count of the marked tear point is set to 1. If there is a difference of the x-coordinate that is less than the width range, the occurrence count of the marked tear point within the width range is increased. If no tear point appears, it is determined whether the occurrence count of the tear point has been updated. If the occurrence count has not been updated, the length of the tear is calculated.

[0061] One of the above technical solutions has the following advantages or beneficial effects: The present invention uses image recognition to intelligently detect the tearing of belts. Compared with the prior art, which determines the degree of belt tearing manually, it can effectively improve the accuracy of measuring belt tearing and provide timely warning of belt breakage. Attached Figure Description

[0062] Figure 1 This is a flowchart of one embodiment of the method of the present invention.

[0063] Figure 2 This is a flowchart of obtaining the tear length in one embodiment of the method of the present invention.

[0064] Figure 3 This is a schematic diagram of selecting a ROI region containing a laser in one embodiment of the present invention.

[0065] Figure 4 This is a schematic diagram of the upper and lower boundaries in one embodiment of the present invention.

[0066] Figure 5 This is a schematic diagram of obtaining the left and right side pixels in one embodiment of the present invention.

[0067] Figure 6 This is a schematic diagram of the structure of one embodiment of the system of the present invention. Detailed Implementation

[0068] Embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0069] In the description of embodiments of the present invention, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the stated features. In the description of embodiments of the present invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0070] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, unless otherwise stated, "a plurality of" means two or more. Those skilled in the art will understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0071] like Figures 1-5 As shown, a method for detecting belt tears based on image recognition includes the following steps:

[0072] Step S1: Use a laser to scan the belt to obtain a running image of the belt, and select a ROI region containing the laser in the running image;

[0073] Step S2: Crop the original image of the ROI region for each frame in the running image;

[0074] Step S3: Fit the ROI region of each frame of the original image to obtain the upper and lower boundaries of the laser.

[0075] Step S4: Obtain the group of suspected tear points by using the height difference between the upper and lower boundaries;

[0076] Step S5: Obtain the curvature of the suspected tear point group and its adjacent pixels, and determine whether the curvature of the suspected tear point group meets the curvature threshold. If it does, then the point is taken as the tear point.

[0077] Step S6: Count the number of times the tear point appears in the running image within the width range, and obtain the tear length of the belt by the running speed of the belt and the number of frames in which the tear point appears.

[0078] To address the problem that existing technologies cannot manually inspect belt tears, this invention proposes a method to replace manual inspection. For example... Figure 3 As shown, a laser-emitting machine is first installed below the transport aircraft. The laser beam is aimed at the conveyor belt, and a camera captures images of the belt in motion, creating a running image. Since the positions of the camera, laser emitter, and conveyor belt are relatively fixed, the laser emitter's position in the running image is also fixed. Therefore, after acquiring the running image, the Region of Interest (ROI) containing the laser beam can be manually selected. Subsequent calculations only require cropping the ROI and processing the image within it, significantly reducing the amount of image processing and thus improving processing speed.

[0079] When a belt is torn, the color of the torn area will change compared to the belt body. Laser scanning makes it easier to distinguish the torn area from the belt body. Then, binarization and other methods can be used to differentiate the torn area from the belt. Since the laser has a certain thickness, after obtaining the Region of Interest (ROI), it needs to be fitted to the ROI to obtain the upper and lower boundaries of the laser beam. Figure 4 As shown, Figure 4 The two black lines in the center represent the upper and lower boundaries, respectively. Under binarization, the torn area and its upper or lower boundary appear as a single unit in the image. Therefore, the height difference between the upper and lower boundaries can be used to determine if a tear point exists. However, in actual use, belts are uneven, and images taken on a camera typically show a wider center and narrower edges. Relying solely on height can easily lead to errors. Therefore, using the height difference between the upper and lower boundaries only identifies suspected tear points. Furthermore, this invention calculates the curvature of these suspected tear points. Curvature is a physical quantity describing the degree of curvature of a curve. On a two-dimensional plane, the curvature of a point on the curve can be calculated to understand the degree of curvature at that point. After a belt tears, due to its material structure, it stretches and deforms during tearing, resulting in a relatively large curvature at the tear point. The curvature value can then be used to further determine whether the height difference between the upper and lower boundaries is caused by the tear itself or by camera error.

[0080] Furthermore, the determination of each tear point is made by analyzing the image material of each frame. When a tear point appears repeatedly within the range, it can be determined that the same belt is torn. Since the belt's running speed is fixed, the length of the tear can be obtained by considering the belt's running speed, the number of frames, and the number of times the tear point appears.

[0081] This invention uses image recognition to intelligently detect belt tears. Compared to existing technologies that rely on manual determination of belt tear extent, this invention effectively improves the accuracy of belt tear measurement and provides timely warnings of belt breakage.

[0082] Preferably, the specific steps of step S3 are as follows:

[0083] Step S1: Use the cv2.medianBlur function to perform median filtering on the original image to remove noise and details while preserving edge information.

[0084] Step S2: Binarize the filtered original image to obtain the approximate outline of the laser in the original image; specifically, select an appropriate threshold for binarization according to the actual situation, and then use the cv2.threshold function to binarize the filtered original image.

[0085] Step A3: Perform region growing or connected region processing on the general outline of the laser to obtain the largest laser area;

[0086] Specifically, when performing region growing or connected region processing, the find_max_region function can be used.

[0087] Step S4: Perform edge detection on the largest laser area to obtain the laser contour; specifically, the cv2.findContours function can be used to implement edge detection.

[0088] Step S5: Use a curve fitting algorithm to obtain the upper and lower boundaries of the laser profile.

[0089] Preferably, the laser profile can be processed by smoothing, interpolation, etc. before the curve fitting algorithm is performed to facilitate the subsequent curve fitting, and the curve fitting algorithm can be the least squares method.

[0090] Preferably, the specific steps of step S4 are as follows:

[0091] Step S41: Traverse the upper and lower boundaries, and store the pixel coordinates of the upper and lower boundaries in two coordinate tables respectively.

[0092] Step S42: Treat two pixels with the same horizontal coordinate in the two coordinate tables as a group of judgment points;

[0093] Step S43: Obtain the difference in the ordinates of two pixels in the judgment point group. If the difference in the ordinates of two pixels is greater than the difference threshold, then set the judgment point group as a suspected tear point group.

[0094] Specifically, after obtaining the upper and lower boundaries, the coordinates of the pixels within each boundary are acquired. These coordinates are then grouped according to their x-coordinates to obtain multiple judgment point groups. The y-coordinates of any two pixels within a judgment point group are subtracted to obtain the difference. If the difference is greater than a threshold, it indicates a relatively large gap between the upper and lower boundaries at that x-coordinate, potentially a tear point. Therefore, the judgment point group with a difference greater than the threshold is designated as a suspected tear point group.

[0095] It is important to note that when calculating the height difference between the upper and lower boundaries, the actual scale relationship between pixels must also be considered. Typically, the conversion relationship between pixels and actual length can be calculated based on the pixel size of the image acquisition device and the actual size of the object in the image. This conversion is then applied to the height difference calculation to ensure that the resulting difference threshold meets the judgment requirements.

[0096] Preferably, the specific steps of step S5 are as follows:

[0097] Step S51: Obtain the adjacent pixels on the left and right sides of two pixels in the suspected tear point group respectively, and obtain the distance t from the two adjacent pixels to the pixels in the suspected tear point group respectively. a and distance t b ;

[0098] like Figure 5 As shown, (x1, y1) is the pixel at the upper boundary of the suspected tear point group, while (x0, y0) and (x2, y2) are the pixels on the left and right sides of the upper boundary pixel in the suspected tear point group, respectively. Since only one boundary at the end of the tear has a curvature greater than a threshold, this invention requires obtaining the curvature of two pixels in the suspected tear point group simultaneously. As long as the curvature of one pixel is greater than the curvature threshold, the suspected tear point group can be identified as a tear point.

[0099] Step S52: Construct a quadratic function model: y = a0 + a1x + a2x 2 ;

[0100] Using distance t a and distance t b The first matrix is ​​constructed as follows:

[0101]

[0102] Step S53: Calculate the specific values ​​of the coefficient vectors a0, a1, and a2 using the first matrix and a quadratic function model;

[0103] Specifically, the coordinates of one pixel in the three suspicious point groups and the pixels on its left and right sides can be used to calculate the results. The formula for solving this problem is as follows: By substituting the corresponding pixel coordinates, the specific values ​​of the coefficient vectors a0, a1, and a2 can be obtained.

[0104] Step S54: Obtain the curvature of the pixels in the suspicious point group using coefficient vectors a1 and a2, where the formula for obtaining the curvature is as follows: Where b1 = a1 + 2a2x2, b2 = 2a2, and x2 is the x-coordinate of one of the adjacent pixels;

[0105] Step S55: Determine whether the curvature of any pixel in the suspicious point group is greater than the curvature threshold. If it is greater, mark the suspicious point group as a tear point.

[0106] Preferably, the specific steps of step S6 are as follows:

[0107] Once a tear point first appears in the current frame, mark the tear point and set the count of the marked tear point to 1.

[0108] Determine if a tear point appears in the next frame. If a tear point appears, obtain the difference between the x-coordinate of the tear point in this frame and the x-coordinates of all marked tear points in the previous frame. Determine if the difference of the x-coordinates is greater than the width range. If the difference of all x-coordinates is greater than the width range, mark the tear point and set the occurrence count of the marked tear point to 1. If there is a difference of x-coordinates that is less than the width range, increase the occurrence count of the marked tear points within the width range. If no tear point appears, determine if the occurrence count of the tear point has been updated. If the occurrence count has not been updated, calculate the length of the tear.

[0109] The formula for calculating the tear length is as follows: S = n * v / e, where n is the number of times the tear point occurs, v is the speed of the belt, and e is the number of frames.

[0110] For example, if tear point A and tear point B appear for the first time in the current frame, and tear point A and tear point B are marked as A1 and A2 respectively, a record table of the occurrence count of A1 and A2 is created, and the occurrence count of A1 and A2 is recorded as 1 respectively;

[0111] The next frame of the image is then identified. If tear point C appears, the x-coordinate of tear point C is subtracted from the x-coordinates of tear points A and B, respectively. The difference is then checked against the width range. If the width range is 3, and the x-coordinates of tear points A and B are 2, 10, and 3 respectively, the difference between their x-coordinates is 1, which is less than the width range of 3. Therefore, tear point C and tear point A are considered to be from the same tear, and tear point C is marked as A1. The occurrence count of A1 is then updated to 2. However, the difference between the x-coordinates of tear points A and C is 7, which is greater than the width range of 3. Therefore, the occurrence count of A2 is not increased, and the tear length of A2 is calculated based on its occurrence count. If the width range is 3, the x-coordinate of tear point A is 15, the x-coordinate of tear point B is 10, and the x-coordinate of tear point C is 3, then the difference between the x-coordinates of tear point A and tear point B and tear point C is greater than the width range. In this case, tear point C is marked as A3, the occurrence number of A3 is recorded as 1, and the tear lengths of A1 and A2 are calculated based on the occurrence numbers of A1 and A2 respectively.

[0112] A system for detecting belt tears based on image recognition, using the aforementioned method for detecting belt tears based on image recognition, includes: a selection module, a cropping module, a fitting module, a judgment module, a determination module, and a length calculation module;

[0113] The selection module is used to scan the belt with a laser to obtain a running image of the belt, and to select a ROI region containing the laser in the running image;

[0114] The cropping module is used to crop out the original image of the ROI region of each frame in the running image;

[0115] The fitting module is used to fit the ROI region of each frame of the original image to obtain the upper and lower boundaries of the laser.

[0116] The judgment module is used to obtain a group of suspected tear points by using the height difference between the upper and lower boundaries;

[0117] The determining module is used to obtain the curvature of the suspected tear point group and its adjacent pixels, and to determine whether the curvature of the suspected tear point group meets the curvature threshold. If it does, the point is taken as the tear point.

[0118] The length calculation module is used to count the number of times the tear point appears in the running image within the width range, and to obtain the tear length of the belt by the running speed of the belt and the number of frames in which the tear point appears.

[0119] Preferably, the fitting module includes a first processing submodule, a second processing submodule, a third processing submodule, a fourth processing submodule, and a fifth processing submodule;

[0120] The first processing submodule is used to perform median filtering on the original image using the cv2.medianBlur function;

[0121] The second processing submodule is used to binarize the filtered original image to obtain the approximate outline of the laser in the original image;

[0122] The three processing submodules are used to perform region growth or connected region processing on the general outline of the laser to obtain the largest laser region.

[0123] The fourth processing submodule is used to perform edge detection on the largest laser area to obtain the laser contour.

[0124] The fifth processing submodule is used to apply a curve fitting algorithm to the laser profile to obtain the upper and lower boundaries of the laser.

[0125] Preferably, the judgment module includes a traversal submodule, a grouping submodule, and a difference calculation submodule;

[0126] The traversal submodule is used to traverse the upper and lower boundaries, and store the pixel coordinates of the upper and lower boundaries after traversal in two coordinate tables respectively.

[0127] The grouping submodule is used to treat two pixels with the same horizontal coordinate in two coordinate tables as a group of judgment points.

[0128] The difference calculation submodule is used to obtain the difference in the ordinate of two pixels in the judgment point group. If the difference in the ordinate of two pixels is greater than the difference threshold, the judgment point group is set as a suspected tear point group.

[0129] Preferably, the determining module includes a first calculation submodule, a second calculation submodule, a third calculation submodule, a fourth calculation submodule, and a fifth calculation submodule;

[0130] The first calculation submodule is used to obtain the adjacent pixels on the left and right sides of two pixels in the suspected tear point group, and to obtain the distance t from the two adjacent pixels to the pixels in the suspected tear point group. a and distance t b ;

[0131] The second calculation submodule is used to construct a quadratic function model: y = a0 + a1x + a2x 2 ;

[0132] Using distance t a and distance t b Construct the first matrix;

[0133] The third calculation submodule is used to calculate the specific values ​​of the coefficient vectors a0, a1 and a2 by using the first matrix and a quadratic function model.

[0134] The fourth calculation submodule is used to obtain the curvature of the pixels in the suspicious point group through coefficient vectors a1 and a2;

[0135] The fifth calculation submodule is used to determine whether the curvature of any pixel in the suspicious point group is greater than the curvature threshold. If it is greater, the suspicious point group is marked as a tear point.

[0136] Preferably, the length calculation module includes a first analysis submodule and a second analysis submodule;

[0137] The method is to mark the tear point after it first appears in the current frame, and set the number of times the marked tear point appears to be 1.

[0138] The second analysis submodule is used to determine whether a tear point appears in the next frame. If a tear point appears, the difference between the x-coordinate of the tear point in this frame and the x-coordinate of all marked tear points in the previous frame is obtained one by one. It is then determined whether the difference of the x-coordinate is greater than the width range. If the difference of all the x-coordinates is greater than the width range, the tear point is marked and the occurrence count of the marked tear point is set to 1. If there is a difference of the x-coordinate that is less than the width range, the occurrence count of the marked tear point within the width range is increased. If no tear point appears, it is determined whether the occurrence count of the tear point has been updated. If the occurrence count has not been updated, the length of the tear is calculated.

[0139] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "illustrative embodiment," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0140] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A method for detecting belt tearing based on image recognition, characterized in that, Includes the following steps: Step S1: Use a laser to scan the belt to obtain a running image of the belt, and select a ROI region containing the laser in the running image; Step S2: Crop the original image of the ROI region for each frame in the running image; Step S3: Fit the ROI region of each frame of the original image to obtain the upper and lower boundaries of the laser. Step S4: Obtain the group of suspected tear points by using the height difference between the upper and lower boundaries; Step S5: Obtain the curvature of the suspected tear point group and its adjacent pixels, and determine whether the curvature of the suspected tear point group meets the curvature threshold. If it does, then the point is taken as the tear point. The specific steps of step S5 are as follows: Step S51: Obtain the adjacent pixels on the left and right sides of each of the two pixels in the suspected tear point group, and obtain the distance from each of the two adjacent pixels to the pixels in the suspected tear point group. and distance ; Step S52: Construct a quadratic function model: ; Use distance and distance The first matrix is ​​constructed as follows: ； Step S53: Calculate the coefficient vector using the first matrix and a quadratic function model. , and The specific value; Step S54: Through the coefficient vector and Obtain the curvature of the pixels in the suspicious point group, where the formula for obtaining the curvature is as follows: ;in , , where x2 is the x-coordinate of one of the adjacent pixels; Step S55: Determine whether the curvature of any pixel in the suspicious point group is greater than the curvature threshold. If it is greater, mark the suspicious point group as a tear point. Step S6: Count the number of times the tear point appears in the running image within the width range, and obtain the tear length of the belt by the running speed of the belt and the number of frames in which the tear point appears; The specific steps of step S6 are as follows: Once a tear point first appears in the current frame, mark the tear point and set the count of the marked tear point to 1. Determine if a tear point appears in the next frame. If a tear point appears, obtain the difference between the x-coordinate of the tear point in this frame and the x-coordinates of all marked tear points in the previous frame. Determine if the difference of the x-coordinates is greater than the width range. If the difference of all x-coordinates is greater than the width range, mark the tear point and set the occurrence count of the marked tear point to 1. If there is a difference of x-coordinates that is less than the width range, increase the occurrence count of the marked tear points within the width range. If no tear point appears, determine if the occurrence count of the tear point has been updated. If the occurrence count has not been updated, calculate the length of the tear. The formula for calculating the tear length is as follows: , where n is the number of tear points, v is the belt speed, and e is the number of frames.

2. The method for detecting belt tearing based on image recognition according to claim 1, characterized in that, The specific steps of step S3 are as follows: Step A1: Use the cv2.medianBlur function to perform median filtering on the original image; Step A2: Binarize the filtered original image to obtain the approximate outline of the laser in the original image; Step A3: Perform region growing or connected region processing on the general outline of the laser to obtain the largest laser area; Step A4: Perform edge detection on the largest laser area to obtain the laser contour; Step A5: Use a curve fitting algorithm to obtain the upper and lower boundaries of the laser profile.

3. The method for detecting belt tearing based on image recognition according to claim 1, characterized in that, The specific steps of step S4 are as follows: Step S41: Traverse the upper and lower boundaries, and store the pixel coordinates of the upper and lower boundaries in two coordinate tables respectively. Step S42: Treat two pixels with the same horizontal coordinate in the two coordinate tables as a group of judgment points; Step S43: Obtain the difference in the ordinates of two pixels in the judgment point group. If the difference in the ordinates of two pixels is greater than the difference threshold, then set the judgment point group as a suspected tear point group.

4. A system for detecting belt tears based on image recognition, using the method for detecting belt tears based on image recognition as described in any one of claims 1 to 3, characterized in that, include: The module includes a selection module, a cropping module, a fitting module, a judgment module, a confirmation module, and a length calculation module. The selection module is used to scan the belt with a laser to obtain a running image of the belt, and to select a ROI region containing the laser in the running image; The cropping module is used to crop out the original image of the ROI region of each frame in the running image; The fitting module is used to fit the ROI region of each frame of the original image to obtain the upper and lower boundaries of the laser. The judgment module is used to obtain a group of suspected tear points by using the height difference between the upper and lower boundaries; The determining module is used to obtain the curvature of the suspected tear point group and its adjacent pixels, and to determine whether the curvature of the suspected tear point group meets the curvature threshold. If it does, the point is taken as the tear point. The determination module includes a first calculation submodule, a second calculation submodule, a third calculation submodule, a fourth calculation submodule, and a fifth calculation submodule; The first calculation submodule is used to obtain the adjacent pixels on the left and right sides of two pixels in the suspected tear point group, and to obtain the distance from the two adjacent pixels to the pixels in the suspected tear point group. and distance ; The second calculation submodule is used to construct a quadratic function model: ; Use distance and distance Construct the first matrix; The third calculation submodule is used to calculate the coefficient vector by combining the first matrix with a quadratic function model. , and The specific value; The fourth calculation submodule is used to process the coefficient vector. and Obtain the curvature of the pixels in the suspicious point group; The fifth calculation submodule is used to determine whether the curvature of any pixel in the suspicious point group is greater than the curvature threshold. If it is greater, the suspicious point group is marked as a tear point. The length calculation module is used to count the number of times the tear point appears in the running image within the width range, and to obtain the tear length of the belt by the running speed of the belt and the number of frames in which the tear point appears; The length calculation module includes a first analysis submodule and a second analysis submodule; The method is to mark the tear point after it first appears in the current frame, and set the number of times the marked tear point appears to be 1. The second analysis submodule is used to determine whether a tear point appears in the next frame. If a tear point appears, the difference between the x-coordinate of the tear point in this frame and the x-coordinate of all marked tear points in the previous frame is obtained one by one. It is then determined whether the difference of the x-coordinate is greater than the width range. If the difference of all the x-coordinates is greater than the width range, the tear point is marked and the occurrence count of the marked tear point is set to 1. If there is a difference of the x-coordinate that is less than the width range, the occurrence count of the marked tear point within the width range is increased. If no tear point appears, it is determined whether the occurrence count of the tear point has been updated. If the occurrence count has not been updated, the length of the tear is calculated.

5. The system for detecting belt tearing based on image recognition according to claim 4, characterized in that, The fitting module includes a first processing submodule, a second processing submodule, a third processing submodule, a fourth processing submodule, and a fifth processing submodule; The first processing submodule is used to perform median filtering on the original image using the cv2.medianBlur function; The second processing submodule is used to binarize the filtered original image to obtain the approximate outline of the laser in the original image; The three processing submodules are used to perform region growth or connected region processing on the general outline of the laser to obtain the largest laser region. The fourth processing submodule is used to perform edge detection on the largest laser area to obtain the laser contour. The fifth processing submodule is used to apply a curve fitting algorithm to the laser profile to obtain the upper and lower boundaries of the laser.

6. The system for detecting belt tearing based on image recognition according to claim 5, characterized in that, The judgment module includes a traversal submodule, a grouping submodule, and a difference calculation submodule; The traversal submodule is used to traverse the upper and lower boundaries, and store the pixel coordinates of the upper and lower boundaries after traversal in two coordinate tables respectively. The grouping submodule is used to treat two pixels with the same horizontal coordinate in two coordinate tables as a group of judgment points. The difference calculation submodule is used to obtain the difference in the ordinate of two pixels in the judgment point group. If the difference in the ordinate of two pixels is greater than the difference threshold, the judgment point group is set as a suspected tear point group.

Images