Belt roller deviation detection method and device, computer device and storage medium

Abstract

The application relates to the field of machine recognition, and discloses a belt roller deviation detection method and device, computer equipment and a storage medium, the method comprises the following steps: acquiring a belt roller image, and detecting all edge line segments in the belt roller image through a line segment detection model; screening all the edge line segments, and determining a left edge line segment, a right edge line segment and a lower edge line segment of the belt roller from the remaining edge line segments; and determining whether the belt roller deviates according to the positions of the left edge line segment, the right edge line segment and the lower edge line segment. The belt roller deviation detection is fully automated, the detection cost is reduced, and the production efficiency is improved.

Description

Technical Field

[0001] This invention relates to the field of machine recognition, and in particular to a method, apparatus, computer equipment, and storage medium for detecting belt roller misalignment. Background Technology

[0002] Currently, there are generally three solutions for tension roller misalignment: 1. Manual inspection and recording; 2. Using a mobile camera (robot) to capture images and then manually judging them in the background; 3. Traditional visual recognition.

[0003] The first two methods require human intervention. The first method requires workers to inspect each item sequentially in multiple production environments, or requires multiple workers, resulting in high labor costs and low efficiency. The second method requires additional hardware costs and also requires constant human inspection, making full automation impossible. The third method uses traditional vision processing methods, but the image quality from cameras varies significantly in different production environments, and traditional vision recognition has poor robustness, making it unsuitable for widespread application across multiple production environments. Summary of the Invention

[0004] In a first aspect, this application provides a method for detecting belt roller misalignment, including:

[0005] Acquire an image of the belt roller and detect all edge line segments in the image of the belt roller using a line segment detection model;

[0006] All edge segments are filtered, and the left, right, and lower edge segments of the belt roller are determined from the remaining edge segments.

[0007] The position of the left edge segment, the right edge segment, and the lower edge segment determines whether the belt roller is offset.

[0008] Furthermore, the filtering of all edge segments includes:

[0009] Based on the probability scores of each edge segment, remove edge segments other than straight lines;

[0010] Then, the area where the belt roller is located is identified by the line segment detection model, and a positioning frame is set according to the area where the belt roller is located, and the edge line segments located outside the positioning frame are screened out.

[0011] Calculate the identity deviation and included angle between the remaining edge segments, and remove overlapping edge segments based on the identity deviation and included angle.

[0012] Furthermore, the step of removing overlapping edge segments based on the identity deviation and the included angle includes:

[0013] Calculate the identity deviation and included angle of the first comparison line segment and the second comparison line segment. When the identity deviation is less than a preset value or the included angle is less than a preset angle, the second comparison line segment is screened out as the overlapping edge line segment of the first comparison line segment.

[0014] The formula for calculating the degree of identity deviation is as follows:

[0015]

[0016] In the formula, J is the degree of identity deviation, L is the length of the first comparison line segment, dp is the maximum distance between any endpoint of the first comparison line segment and any endpoint of the second comparison line segment, θ is the angle between the first comparison line segment and the second comparison line segment, and dl is the distance from the center of the second comparison line segment to the first comparison line segment.

[0017] Furthermore, the step of setting the positioning frame according to the area where the belt roller is located includes:

[0018] The belt roller image is divided into grids, and the target confidence, center coordinates, and length and width of the positioning box are calculated for each grid in turn.

[0019] The position of the belt roller in the belt roller image is determined by the target confidence of each grid, and the center coordinates and the length and width of the positioning frame are determined to determine the range of the positioning frame.

[0020] Furthermore, the line segment detection model is constructed by adding a detection branch before the jump connection layer in the lightweight real-time detection model. The detection branch includes multiple recognition modules and is used to determine the position of the positioning box.

[0021] Furthermore, determining the left edge segment, right edge segment, and lower edge segment of the belt roller from the remaining edge segments includes:

[0022] Calculate the relative positional relationship between any two edge segments, filter out the edge segment group with two parallel edge segments and perpendicular to the same edge segment, and filter the edge segment group according to the preset edge segment length to determine the left edge segment, right edge segment and lower edge segment of the belt roller.

[0023] Furthermore, determining whether the belt roller is offset based on the positions of the left edge segment, the right edge segment, and the lower edge segment includes:

[0024] The first intersection point of the left edge segment and the lower edge segment, and the second intersection point of the right edge segment and the lower edge segment are determined in real time.

[0025] If the first intersection point exceeds the preset left limit, or the second intersection point exceeds the preset right limit, or the lower edge line segment intersects with the preset upper and lower alarm lines, then the belt roller is determined to be offset.

[0026] Secondly, this application also provides a belt roller misalignment detection device, comprising:

[0027] The recognition module is used to acquire the belt roller image and detect all edge line segments in the belt roller image using a line segment detection model;

[0028] The filtering module is used to filter all edge segments and determine the left edge segment, right edge segment and lower edge segment of the belt roller from the remaining edge segments;

[0029] The detection module is used to determine whether the belt roller is offset based on the positions of the left edge line segment, the right edge line segment, and the lower edge line segment.

[0030] Thirdly, this application also provides a computer device, including a processor and a memory, wherein the memory stores a computer program, and the computer program executes the belt roller misalignment detection method described in any one of the claims when it is run on the processor.

[0031] Fourthly, this application also provides a readable storage medium storing a computer program that, when run on a processor, executes any of the belt roller misalignment detection methods described in the present application.

[0032] This invention relates to the field of machine recognition and discloses a method, apparatus, computer equipment, and storage medium for detecting belt roller misalignment. The method includes: acquiring an image of the belt roller and detecting all edge segments in the image using a line segment detection model; filtering all edge segments and determining the left, right, and bottom edge segments of the belt roller from the remaining edge segments; and determining whether the belt roller is misaligned based on the positions of the left, right, and bottom edge segments. This fully automates belt roller misalignment detection, reduces detection costs, and improves production efficiency. Attached Figure Description

[0033] To more clearly illustrate the technical solution of the present invention, the accompanying drawings used in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope of protection of the present invention. In the various drawings, similar components are numbered similarly.

[0034] Figure 1 This paper illustrates a schematic flowchart of a belt roller misalignment detection method according to an embodiment of this application.

[0035] Figure 2 This paper shows a schematic diagram of a line segment detection model structure according to an embodiment of the present application;

[0036] Figure 3 This illustration shows a schematic diagram of a line segment overlap according to an embodiment of this application;

[0037] Figure 4 This paper illustrates a schematic diagram of belt and roller misalignment detection according to an embodiment of the present application.

[0038] Figure 5 A schematic diagram of a belt roller misalignment detection device according to an embodiment of this application is shown. Detailed Implementation

[0039] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0040] The components of the embodiments of the invention described and illustrated herein can typically be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0041] In the following, the terms “comprising,” “having,” and their cognates, which may be used in various embodiments of the invention, are intended only to indicate a particular feature, number, step, operation, element, component, or combination thereof, and should not be construed as excluding, firstly, the presence of one or more other features, numbers, steps, operations, elements, components, or combinations thereof, or adding the possibility of one or more features, numbers, steps, operations, elements, components, or combinations thereof.

[0042] Furthermore, the terms "first," "second," and "third" are used only to distinguish descriptions and should not be interpreted as indicating or implying relative importance.

[0043] Unless otherwise specified, all terms used herein (including technical and scientific terms) shall have the same meaning as commonly understood by one of ordinary skill in the art to which the various embodiments of the invention pertain. Terms (such as those defined in commonly used dictionaries) shall be interpreted as having the same meaning as in their contextual meaning in the relevant technical field and shall not be interpreted as having an idealized or overly formal meaning, unless clearly defined in the various embodiments of the invention.

[0044] This application is used to identify belt misalignment in tensioning rollers. For tensioning rollers, the belt wound around the rollers needs to be wound correctly and without deviation, otherwise it will affect the winding result of the belt. This application takes real-time pictures of the tensioning rollers with a camera, then uses a line recognition model to detect the edge lines in the image, and then identifies the position of the tensioning rollers. Finally, based on the identified position of the tensioning rollers, it determines whether the belt is misaligned, thus achieving the effect of automatic identification.

[0045] The technical solution of this application will now be described with reference to specific embodiments.

[0046] Example 1

[0047] like Figure 1 As shown, the belt roller misalignment detection method of this embodiment includes:

[0048] Step S100: Obtain an image of the belt roller and detect all edge lines in the image of the belt roller using a line segment detection model.

[0049] A belt roller is a device that consists of a belt wound around its base and driven by a motor to rotate, causing the belt to wind and move. This can be a common example such as a conveyor belt. It's understandable that if the belt or roller in this type of structure is misaligned, it will cause significant disruption and affect production; therefore, it's necessary to detect the degree of misalignment. For this purpose, this embodiment uses edge line segment detection to position the belt roller in the scene.

[0050] In this embodiment, an image of the belt pulley is acquired using a camera. During shooting, the camera should be positioned roughly horizontally and directly facing the pulley. The pulley should be centered as much as possible for clear focus; simultaneously, the image itself should be magnified as much as possible to ensure details are visible, including the complete left, right, and bottom edges of the taut pulley. This results in a clear image. It is understood that the above requirements can be achieved by adjusting the camera's shooting position, angle, and lens focal length.

[0051] After obtaining the image, this embodiment uses a line segment detection model for identification. This model is an extension of the M-LSD (Lightweight Real-Time Detection) model, specifically adding a detection branch to the open-source M-LSD model for detecting the tension belt device. The main extensions are as follows:

[0052] like Figure 2As shown, the top is the complete M-LSD model, and the bottom dashed box contains the newly added detection branch. In this embodiment, the detection branch includes 6 blocks (recognition modules), with the last block connected to a 1×1 convolutional layer, followed by an output layer with 5 channels (regression values ​​of confidence and 4 localization boxes). The confidence level represents the probability of the target image existing in the current grid. In this embodiment, the target image is a belt roller.

[0053] In addition to the detection branch, the above model also incorporates mosaic image enhancement during training. Therefore, the loss function for the detection branch is divided into classification loss and regression loss. Focal loss is used for classification loss, and GIoU is used for regression loss.

[0054] The two specific loss functions are as follows:

[0055]

[0056]

[0057] L det =f cls *L cls +f reg *L reg ;

[0058] After merging the loss functions of the original M-LSD model, the total loss is:

[0059] L total =L tp +L sol +L geo +f det *L det ;

[0060] In the above formula, α represents the positive and negative samples of focal loss, γ represents the modulation factor of easy and difficult samples, p represents the classification output probability, IOU represents the intersection-union ratio of the predicted box and the label box, A represents the minimum convex closed box of the predicted box and the label box, and U represents the union of the predicted box and the label box.

[0061] L clst L represents the classification loss of the detection branch. reg For regression loss, L de To detect the total loss f cls f reg f det These represent the weighting coefficients for the corresponding losses. L tp , where L is the total loss represented by line segment TP in the original M-LSD model. sol To enhance the TP representation loss, L geoThe loss function is geometric. The model obtained by training based on the above loss function is the line segment detection model used in this embodiment.

[0062] Specifically, during the recognition process, the image to be recognized is segmented by a network. The segmentation network divides the image into grid images, and then each grid is recognized. For each grid, the corresponding confidence score and bounding box regression parameters are output. In this way, the grid range with belt rollers can be identified, thereby determining the range of the bounding box.

[0063] Assuming the center of the bounding box is located at the (i, j) grid in the output layer, then in this embodiment, the bounding box regression equation is:

[0064] C x = j + sigmoid(dx) * 2 - 0.5;

[0065] C y = i + sigmoid(dy) * 2 - 0.5;

[0066] W = e w ;

[0067] H = e h ;

[0068] In the formula, C x C represents the x-coordinate of the center point of the positioning box at the feature layer scale. y Let W be the ordinate of the center point of the bounding box at the feature layer scale, W be the width of the bounding box, H be the height of the bounding box, and sigmoid be the activation function of the neural network.

[0069] The value range of sigmoid(dy)*2-0.5 is -0.5 to 1.5. With this design, the predictable center position of each grid point in the grid diagram can be extended to the adjacent grid points, so the number of positive samples of the location box can be increased to about three times the original.

[0070] At the same time, it will also identify all edge segments in the image. Identifying edge segments is also to locate the belt roller. However, in the captured image, other edge segments will inevitably be identified. Therefore, after identifying the edge segments, a positioning frame is set based on the grid range where the roller and belt are located. Based on the location of the positioning frame, all edge segments outside the positioning frame are screened out to remove these interference items.

[0071] At the same time, the identified edge segments are [(L n x1,L n y1),(L n x2,L n y2),(L n xc ,L n y c ),S n The shape of ]], (L n x1,L n y1),(L n x2,L n y2) represents the coordinates of the two endpoints of the nth line segment. (L) n x c ,L n y c S represents the center of the nth line segment. n The probability score represents the probability of a line segment being a straight line, ranging from 0 to 1. It indicates the probability that the line segment is a straight line. Since the belt roller is rectangular, the curve must not be an edge segment on the belt roller. Therefore, based on the probability score of the line segment, a batch of line segments that are not straight lines can be filtered out.

[0072] Therefore, after the dual screening of line segment probability scores and the aforementioned positioning frame, the remaining line segments are roughly those related to the belt roller.

[0073] Step S200: Filter all edge segments and determine the left edge segment, right edge segment and lower edge segment of the belt roller from the remaining edge segments.

[0074] It is understandable that after the recognition model performs edge segment recognition, it may identify multiple overlapping line segments around the same line segment. For example, because there is a fold at the lower edge of the belt roller, it is easy to identify multiple parallel and adjacent edge line segments. Only one of these edge line segments needs to be kept. Therefore, it is necessary to further filter these line segments to remove overlapping line segments.

[0075] Therefore, the remaining line segments will be compared pairwise. For the two line segments being compared, this embodiment will calculate the identity deviation of the two line segments and determine whether the two line segments overlap based on the identity deviation.

[0076] like Figure 3 As shown, the upper line segment is the first comparison line segment with length L, and the lower line segment is the second comparison line segment. The expression for calculating the degree of deviation of the identity between these two line segments is:

[0077]

[0078] In the formula, J is the degree of identity deviation, L is the length of the first comparison line segment, dp is the maximum distance between any endpoint of the first comparison line segment and any endpoint of the second comparison line segment, θ is the angle between the first comparison line segment and the second comparison line segment, and dl is the distance from the center of the second comparison line segment to the first comparison line segment.

[0079] If J is less than a preset value, such as 1, or θ < 0.175 (10 degrees in degrees), then the two line segments are considered to overlap, and one of them can be discarded.

[0080] After deduplication, there may still be redundant edge segments. This is understandable. In order to express the position and orientation of the belt roller, this embodiment only needs to obtain the left edge segment, right edge segment, and bottom edge segment of the belt roller. From the geometric relationship of these three segments, we know that they are two parallel segments in the same direction perpendicular to the same line segment. The edge segments identified above are equivalent to vectors in the figure. Therefore, the dot product of any pair of vectors is calculated to obtain an m×m matrix M. Assuming LB is the bottom edge segment, LR is the right edge segment, and LL is the left edge segment, with i, j, k in the segment group, then M satisfies ij M ik Both are approximately 0. At the same time, the left and right edge segments should be of similar length and greater than the lower edge segment. Then, further filtering is performed based on the segment length to find the target segment group [LL,LR,LB].

[0081] In summary, after initial identification of multiple edge segments, a rough screening is performed using segment probability scores and bounding boxes to remove irrelevant edge segments. Then, a screening is performed based on the degree of identity deviation to ensure that no overlapping segments remain. Finally, a final screening is performed based on the geometric characteristics of the required left, right, and lower edge segments to find these three segments.

[0082] Step S300: Determine whether the belt roller is offset based on the positions of the left edge segment, the right edge segment, and the lower edge segment.

[0083] It is understandable that once the above three line segments are determined, it is equivalent to accurately positioning the belt roller. Therefore, based on these three line segments, it is possible to determine whether the belt roller is misaligned.

[0084] Specifically, such as Figure 4 As shown, the positions of the belt 100 and the roller 200 are defined by the left edge segment, the right edge segment, and the lower edge segment. To determine whether the belt or roller is misaligned, an upper alarm line 300, a lower alarm line 600, a left limit switch 500, and a right limit switch 400 are set. The dotted line represents the theoretical position of the lower edge segment.

[0085] The alarm lines and limit blocks mentioned above are designed based on the parameters of the belt roller itself. The presence or absence of belt roller deviation is determined by the alarm lines and limit blocks.

[0086] Specifically, the first intersection point of the left edge segment and the lower edge segment, and the second intersection point of the right edge segment and the lower edge segment can be determined in real time.

[0087] If the first intersection point exceeds the preset left limit of 500, or the second intersection point exceeds the preset right limit of 400, or the lower edge line segment intersects with the preset upper and lower alarm lines, then the belt roller is determined to be offset.

[0088] Understandable. Figure 4 The upper alarm line 300 and lower alarm line 600 can limit the offset angle and vertical displacement of the belt roller to a certain extent, preventing the belt roller from tilting too much. At the same time, the left limit 500 and right limit 400 can prevent the belt roller from shifting to the left or right. Thus, the belt roller is confined within a certain space. When the first intersection point, the second intersection point, or the lower edge line segment exceeds the area defined by the upper alarm line 300, lower alarm line 600, left limit 500, and right limit 400, an alarm will be triggered, thereby monitoring the belt roller.

[0089] It is understandable that the above detection methods can be applied to terminal control equipment as well as inspection robots, making the detection method more automated and improving detection efficiency.

[0090] The belt roller misalignment detection method in this embodiment uses a camera to capture images, then identifies the required left, right, and bottom edge segments, and uses these three segments to determine if the belt roller is misaligned. This misalignment detection method is fully automated, requiring no manual operation, greatly reducing labor costs, and increasing the precision of the identification, resulting in more accurate identification results and faster detection of belt roller misalignment, thereby increasing production efficiency.

[0091] Example 2

[0092] like Figure 5 As shown, this application also provides a belt roller misalignment detection device, comprising:

[0093] The recognition module 10 is used to acquire the belt roller image and detect all edge line segments in the belt roller image using a line segment detection model.

[0094] The filtering module 20 is used to filter all edge segments and determine the left edge segment, right edge segment and lower edge segment of the belt roller from the remaining edge segments;

[0095] The detection module 30 is used to determine whether the belt roller is offset based on the positions of the left edge line segment, the right edge line segment and the lower edge line segment.

[0096] This application also provides a computer device, including a processor and a memory, wherein the memory stores a computer program, and the computer program, when run on the processor, executes any of the belt roller misalignment detection methods described above.

[0097] This application also provides a readable storage medium storing a computer program that, when run on a processor, executes any of the belt roller misalignment detection methods described above. The method includes: acquiring a belt roller image and detecting all edge segments in the belt roller image using a line segment detection model; filtering all edge segments and determining the left, right, and bottom edge segments of the belt roller from the remaining edge segments; and determining whether the belt roller is misaligned based on the positions of the left, right, and bottom edge segments. This fully automates belt roller misalignment detection, reduces detection costs, and improves production efficiency.

[0098] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can also be implemented in other ways. The apparatus embodiments described above are merely illustrative; for example, the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that, as an alternative implementation, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagram and / or flowchart, and combinations of blocks in the block diagram and / or flowchart, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0099] In addition, the functional modules or units in the various embodiments of the present invention can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0100] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a smartphone, personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0101] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A belt drum run-off detection method, characterized in that, The method comprises the following steps: acquiring a belt roller image and detecting all edge lines in the belt roller image through a line segment detection model; screening all edge lines and determining left edge lines, right edge lines and lower edge lines of the belt roller from the remaining edge lines; determining whether the belt roller is deviated according to positions of the left edge lines, the right edge lines and the lower edge lines; the screening of all edge lines comprises: eliminating edge lines other than straight lines according to probability scores of each edge line; then identifying a region where the belt roller is located through the line segment detection model, setting a positioning frame according to the region where the belt roller is located, and screening edge lines located outside the positioning frame; calculating identity deviation and included angle between remaining each edge line, and removing overlapping edge lines according to the identity deviation and the included angle; the removing of overlapping edge lines according to the identity deviation and the included angle comprises: calculating identity deviation and included angle of a first contrast line and a second contrast line, and removing the second contrast line as an overlapping edge line of the first contrast line when the identity deviation is less than a preset value or the included angle is less than a preset angle; the formula of the identity deviation is: ； wherein, J is the identity deviation, L is the length of the first contrast line, dp is the maximum distance between any end point of the first contrast line and any end point of the second contrast line, θ is the included angle between the first contrast line and the second contrast line, and dl is the distance from the center of the second contrast line to the first contrast line; the line segment detection model is obtained by adding a detection branch before a jump connection layer in a lightweight real-time detection model, wherein the detection branch comprises a plurality of identification modules, and the detection branch is used to determine the position of the positioning frame; the determination of whether the belt roller is deviated according to the positions of the left edge lines, the right edge lines and the lower edge lines comprises: real-time determining a first intersection point of the left edge lines and the lower edge lines, and a second intersection point of the right edge lines and the lower edge lines; if the first intersection point crosses a preset left limit, or the second intersection point crosses a preset right limit, or the lower edge lines intersect with a preset upper and lower alarm line, it is determined that the belt roller is deviated.

2. The belt drum run off detection method according to claim 1, characterized in that the setting of the positioning frame according to the region where the belt roller is located comprises: dividing the belt roller image into grids, and sequentially calculating target confidence, center coordinates and length and width of the positioning frame of each grid; determining the position of the belt roller in the belt roller image through the target confidence of each grid, and determining the center coordinates and the length and width of the positioning frame to determine the range of the positioning frame.

3. The belt run off detection method of claim 1, wherein the determination of the left edge lines, the right edge lines and the lower edge lines of the belt roller from the remaining edge lines comprises: The relative positional relationship between any two edge line segments is calculated, edge line segment groups of two parallel edge line segments and perpendicular to the same edge line segment are screened out, and the edge line segment groups are filtered according to a preset edge line segment length, so as to determine the left edge line segment, the right edge line segment and the lower edge line segment of the belt roller.

4. A belt roller deviation detecting device characterized by comprising: Comprise: The identification module is used for acquiring a belt roller image and detecting all edge line segments in the belt roller image through a line segment detection model; The screening module is used for screening all edge line segments and determining the left edge line segment, the right edge line segment and the lower edge line segment of the belt roller from the remaining edge line segments; The detection module is used for determining whether the belt roller is deviated according to the positions of the left edge line segment, the right edge line segment and the lower edge line segment; The screening of all edge line segments comprises: According to the probability score of each edge line segment, edge line segments other than straight lines are removed; Then the line segment detection model is used to identify the area where the belt roller is located, a positioning frame is set according to the area where the belt roller is located, and edge line segments located outside the positioning frame are screened out; The identity deviation and the included angle between the remaining edge line segments are calculated, and overlapping edge line segments are removed according to the identity deviation and the included angle; The removal of overlapping edge line segments according to the identity deviation and the included angle comprises: The identity deviation and the included angle of a first contrast line segment and a second contrast line segment are calculated, and when the identity deviation is less than a preset value or the included angle is less than a preset angle, the second contrast line segment is screened out as an overlapping edge line segment of the first contrast line segment; The formula for calculating the identity deviation is: ； In the formula, J is the identity deviation, L is the length of the first contrast line segment, dp is the maximum distance between any end point of the first contrast line segment and any end point of the second contrast line segment, θ is the included angle between the first contrast line segment and the second contrast line segment, and dl is the distance from the center of the second contrast line segment to the first contrast line segment; The line segment detection model is obtained by adding a detection branch before the jump connection layer in a lightweight real-time detection model, wherein the detection branch comprises a plurality of identification modules, and the detection branch is used to determine the position of the positioning frame; The determination of whether the belt roller is deviated according to the positions of the left edge line segment, the right edge line segment and the lower edge line segment comprises: The first intersection point of the left edge line segment and the lower edge line segment and the second intersection point of the right edge line segment and the lower edge line segment are determined in real time; If the first intersection point exceeds a preset left limit, or the second intersection point exceeds a preset right limit, or the lower edge line segment intersects with a preset upper and lower alarm line, it is determined that the belt roller is deviated.

5. A computer device, comprising: The device comprises a processor and a memory, the memory stores a computer program, and the computer program executes the belt roller deviation detection method in any one of claims 1 to 3 when running on the processor.

6. A readable storage medium characterized by, The computer program is stored in the memory and is configured to perform the belt run-off detection method according to any one of claims 1 to 3 when running on the processor.

Images