Belt tearing detection method and device based on three-dimensional vision

Abstract

The invention provides a belt tearing detection method and device based on three-dimensional vision. The method comprises the steps of calculating a spatial distance value between each target point and an adjacent point in three-dimensional point cloud data in a direction perpendicular to a tangential direction of a laser arc, screening points of which the spatial distance values exceed a preset distance threshold value, forming an initial candidate point set, starting from each point, expanding a space through a breadth-first search algorithm, and obtaining a target candidate point set, calculating the length and depth of each connected domain; and taking the connected domain meeting the threshold condition as a corresponding potential crack, forming a potential crack set by the spatial position information of all the potential cracks, and generating a belt tearing detection result containing the spatial position information of the real crack and the corresponding alarm level. According to the technical scheme provided by the invention, high-precision dynamic identification and alarm of the surface crack of the belt are realized, the false detection rate is effectively reduced, and the industrial safety monitoring efficiency is improved.

Description

Technical Field

[0001] The present application relates to the technical field of belt anomaly detection, and in particular to a belt tear detection method and device based on three-dimensional vision. Background Art

[0002] In steel production, material conveyor belts (i.e. belts) tear from time to time. Once a longitudinal tear occurs, the conveyor belt will be completely destroyed in a very short time, causing huge economic losses. Even if it can be repaired, it will take considerable manpower and time, which will have a great impact on normal production. In recent years, the use of material conveyor belts has increased, and its application scope has become wider and wider, but so far there is no ideal, mature method and equipment that can be widely promoted to effectively detect longitudinal tearing.

[0003] The existing solution is based on two-dimensional vision technology to achieve automatic detection of longitudinal tears in material conveyor belts, which can replace manual inspection, avoid potential personnel safety accidents, and improve the safe operation level of conveyor belts.

[0004] However, the existing solutions based on two-dimensional visual methods are limited to the extraction of planar information under a single perspective, and it is difficult to effectively distinguish the three-dimensional morphological differences between the interference of stains and shadows on the belt surface and the real cracks. For example, when there is reflection or dust on the belt surface, the two-dimensional image features are prone to produce false edges, resulting in an increase in the false detection rate; at the same time, due to the lack of depth information, the actual depth of the crack cannot be accurately quantified, and the potential risk of small but deep tearing may be missed. In addition, the two-dimensional threshold setting is sensitive to lighting conditions and lacks robustness in complex industrial environments, requiring frequent manual calibration of parameters to maintain detection accuracy. Summary of the invention

[0005] The present application provides a belt tear detection method and device based on three-dimensional vision, which is used to solve the problem of low belt tear detection accuracy in the prior art caused by difficulty in effectively distinguishing real cracks from surface interference, inability to quantify crack depth and insufficient environmental robustness.

[0006] In a first aspect, the present application provides a belt tear detection method based on three-dimensional vision, comprising: The belt surface image is collected by a line laser and an industrial camera deployed under the belt, and the three-dimensional point cloud data of the laser arc in the belt surface image is calculated; Calculating the spatial distance value between each target point and the adjacent points in the three-dimensional point cloud data in the direction perpendicular to the tangent of the laser arc, and screening the points whose spatial distance value exceeds the preset distance threshold to form an initial candidate point set; Starting from each point in the initial candidate point set, the Breadth-First Search (BFS) algorithm is used to expand the points in the space adjacent to each other and belonging to the initial candidate point set to obtain the target candidate point set, so as to form multiple disconnected connected domains, and calculate the length and depth of each connected domain; Filtering the connected domains that do not meet the threshold condition among all connected domains, retaining the connected domains that meet the threshold condition, taking all the connected domains that meet the threshold condition as corresponding potential cracks, and forming a potential crack set with the spatial position information of all the potential cracks, wherein the non-meeting threshold condition means that both the length and the depth do not reach the corresponding preset threshold; During the belt circulation process, if there is a real crack, a belt tear detection result including the spatial position information of the real crack and the corresponding alarm level is generated. The real crack is a potential crack recorded at the same spatial position in the potential crack set in multiple consecutive operation cycles.

[0007] Optionally, starting from each point in the initial candidate point set, expanding points that are spatially adjacent and belong to the initial candidate point set by a breadth-first search algorithm to obtain a target candidate point set includes: Traversing each target point in the initial candidate point set, marking the target point as a member of the current connected domain, and inserting it into a preset queue to be viewed, and calculating the dynamic neighborhood range of the current target point according to the belt running direction and belt running speed; Searching for adjacent points in the dynamic neighborhood that are spatially adjacent and belong to the initial candidate point set along a direction perpendicular to the tangent of the laser arc; Determine whether the neighboring points of the target point belong to the initial candidate point set and are not marked; If yes, mark the adjacent point as a member of the current connected domain, insert it to the end of the preset queue to be checked, take the adjacent point as a new target point, and repeat the steps of marking, inserting, calculating, searching, and judging until all the target points in the preset queue to be checked form a connected domain, clear the preset queue to be checked, and repeat the process of forming a new connected domain until each target point in the initial candidate point set is marked; All target points that are not in any connected domain in the initial candidate point set are deleted to obtain a target candidate point set.

[0008] Optionally, the judgment condition of spatial proximity is that the three-dimensional space Euclidean distance between the adjacent point and the target point is less than a preset connection distance threshold, and the preset connection distance threshold is dynamically set according to the mapping relationship between the spot diameter of the line laser and the belt running speed calibrated by the experiment; The adjacent points in the search space and belonging to the initial candidate point set include: Calculate the tangent direction vector of the laser arc at the target point; Generate a vector perpendicular to the tangent direction of the laser arc according to the tangent direction vector of the laser arc and the unit vector of the belt thickness direction; The preset connected distance threshold is extended in the positive and negative directions of the tangent direction vector of the laser arc to form a strip-shaped search area, and points other than the target point in the strip-shaped search area are determined as adjacent points.

[0009] Optionally, the calculating of the spatial distance value between each target point and the adjacent point in the three-dimensional point cloud data in the direction perpendicular to the tangent of the laser arc, and screening the points whose spatial distance value exceeds a preset distance threshold to form an initial candidate point set includes: The cross-sectional data of the multi-layer composite structure is obtained through the preset belt cross-sectional scanning device, and the correlation model between the belt material and the belt thickness is established; According to the association model, dynamically generate a hierarchical distance threshold for each hierarchical structure where the target point is located; Calculate the spatial distance difference between the target point and the adjacent points in the anisotropic neighborhood in the belt thickness direction; compare the spatial distance difference with the stratified distance threshold, screen out the abnormal points whose spatial distance difference exceeds the stratified distance threshold, and determine the maximum penetration depth and stratified position information of each abnormal point; Constructing a density field based on the spatial distribution characteristics of the abnormal points, and calculating the spatial concentration of each abnormal point in the density field within a preset volume element; The product of the spatial concentration and the maximum penetration depth is determined as a comprehensive anomaly score, and abnormal points whose comprehensive anomaly scores exceed a preset score threshold are screened to form an initial candidate point set.

[0010] Optionally, determining the maximum penetration depth and layered position information of each abnormal point includes: According to the spatial position information of the abnormal point, combined with the geometric model of the layer boundary, the maximum deviation between each abnormal point and the layer boundary is calculated to generate the maximum penetration depth; In combination with the hierarchical index table, hierarchical position analysis is performed on the spatial position information of the abnormal point to obtain hierarchical position information.

[0011] Optionally, forming a potential crack set from the spatial location information of all potential cracks includes: Extracting crack feature points according to the spatial positions of all the potential cracks and forming an initial potential crack point set; Performing spatial clustering on the initial potential fracture point set, and merging adjacent points to form potential fracture segments; Calculating the geometric features of all the potential fracture segments, analyzing the adjacency and intersection relationships between the fracture segments, and constructing a fracture topology map; Based on the geometric continuity and the fracture topology map, global optimization is performed to eliminate low-confidence fracture segments and obtain a potential fracture set.

[0012] Optionally, calculating the length and depth of each connected domain includes: Parsing the spatial positions and boundary information of all the connected domains to generate an initial description information set for each of the connected domains; According to the initial description information set and the preset spatial geometric relationship model, performing boundary extraction for each of the connected domains to obtain coordinates of all boundary points of each connected domain; According to the coordinates of all the boundary points of each of the connected domains, the relative position relationship between all the boundary points of the connected domains is analyzed, two first endpoints corresponding to the longest path are determined, and the distance value between the two first endpoints is used as the length of the connected domain; According to the coordinates of all boundary points of each connected domain, the distance distribution between all boundary points of the connected domain and the preset connected domain center point is analyzed, two second endpoints perpendicular to the longest path are determined, and the distance value between the two second endpoints is used as the depth of the connected domain.

[0013] In a second aspect, the present application provides a belt tear detection device based on three-dimensional vision, comprising: An acquisition module is used to acquire the belt surface image through a line laser and an industrial camera deployed under the belt, and calculate the three-dimensional point cloud data of the laser arc in the belt surface image; A screening module, used to calculate the spatial distance value between each target point and the adjacent points in the three-dimensional point cloud data in the direction perpendicular to the tangent of the laser arc, and screen the points whose spatial distance value exceeds a preset distance threshold to form an initial candidate point set; A calculation module is used to start from each point in the initial candidate point set, expand points that are adjacent in space and belong to the initial candidate point set through a breadth-first search algorithm, obtain a target candidate point set, so as to form multiple disconnected connected domains, and calculate the length and depth of each connected domain; A filtering module is used to filter the connected domains that do not meet the threshold condition in all connected domains, retain the connected domains that meet the threshold condition, take all the connected domains that meet the threshold condition as corresponding potential cracks, and form a potential crack set with the spatial position information of all potential cracks, wherein the non-meeting threshold condition means that both the length and the depth do not reach the corresponding preset threshold; A generation module is used to generate a belt tear detection result including spatial position information of the real crack and a corresponding alarm level if a real crack exists during the belt circulation operation. The real crack is a potential crack recorded at the same spatial position in the potential crack set in multiple consecutive operation cycles.

[0014] In a third aspect, the present application provides a computing device comprising a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a three-dimensional vision-based belt tear detection method as described in any one of the first aspects.

[0015] In a fourth aspect, the present application provides a computer storage medium storing a computer program, which, when executed by a computer, implements a belt tear detection method based on three-dimensional vision as described in any one of the first aspects.

[0016] In an embodiment of the present application, a belt tear detection method based on three-dimensional vision is provided, the method comprising: collecting a belt surface image by a line laser and an industrial camera deployed under the belt, and calculating three-dimensional point cloud data of a laser arc in the belt surface image; calculating the spatial distance value between each target point and an adjacent point in the three-dimensional point cloud data in a direction perpendicular to the tangent of the laser arc, screening points whose spatial distance value exceeds a preset distance threshold, and forming an initial candidate point set; starting from each point in the initial candidate point set, expanding points that are spatially adjacent and belong to the initial candidate point set by a breadth-first search algorithm, and obtaining a target candidate point set, so as to form a plurality of disconnected connected domains , and calculate the length and depth of each connected domain; filter out the connected domains that do not meet the threshold conditions in all connected domains, retain the connected domains that meet the threshold conditions, take all the connected domains that meet the threshold conditions as corresponding potential cracks, and form a potential crack set with the spatial position information of all potential cracks, wherein the threshold condition is not met when the length and depth do not reach the corresponding preset threshold; during the belt circulation process, if there is a real crack, generate a belt tear detection result containing the spatial position information of the real crack and the corresponding alarm level, wherein the real crack is a potential crack recorded at the same spatial position in the potential crack set in multiple consecutive operation cycles.

[0017] This application is based on the three-dimensional point cloud data of laser arcs collected by line lasers and industrial cameras, breaking through the dependence of traditional two-dimensional visual technology on plane information, and can accurately characterize the depth changes of the belt surface morphology, and improve the detection sensitivity of the three-dimensional spatial features of tiny cracks (especially longitudinal tearing). By calculating the spatial distance between adjacent points in the three-dimensional point cloud in the direction of the vertical laser arc tangent, and combining the preset distance threshold to screen candidate points, it can effectively distinguish between real cracks and surface stains, reflections and other plane interferences, reduce the false detection rate caused by environmental noise, and improve the accuracy of belt tear detection results. The breadth-first search algorithm is used to generate interconnected domains that are not connected to each other, and combined with the length and depth double threshold filtering, the pseudo-crack areas formed by isolated noise points are accurately eliminated; at the same time, the potential cracks in the same spatial position are verified through multi-cycle continuous detection to avoid false alarms caused by instantaneous interference, and ensure the reliability and stability of the detection results. By generating crack spatial position and alarm level information in real time, accurate decision support is provided for belt operation and maintenance, meeting the needs of unmanned and high-efficiency tear risk warning in industrial scenarios.

[0018] Furthermore, this method realizes accurate extraction of crack connected domains through dynamic neighborhood range calculation and breadth-first search algorithm. Specifically: first, the dynamic neighborhood range of the target point is determined based on the running direction and speed of the belt, and the spatial adjacent points that meet the three-dimensional Euclidean distance threshold are searched along the direction perpendicular to the tangent of the laser arc; through iterative marking, queue insertion and connected domain expansion process, combined with the strip search area limitation, the initial candidate points are clustered into disconnected connected domains, and finally isolated points are eliminated to form a set of target candidate points. This scheme solves the problem of regional breakage or over-connection caused by belt movement speed changes in traditional connected domain search through dynamic neighborhood and adaptive connected distance threshold design, ensuring the integrity and continuity of crack morphology; combining strip search strategy with three-dimensional Euclidean distance constraint, accurately distinguishing real cracks from discrete noise points, avoiding misjudgment caused by local point cloud anomalies; through parameter mapping of spot diameter and belt speed, the search range is dynamically optimized, the algorithm's adaptability to complex working conditions is improved, and the recognition robustness of the crack space topological structure is enhanced.

[0019] These and other aspects of the present application will become more clearly understood in the description of the following embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0021] Figure 1A flowchart of a belt tear detection method based on three-dimensional vision provided in an embodiment of the present application; Figure 2 A schematic diagram of the positions of a line laser and a camera provided in an embodiment of the present application; Figure 3 A schematic diagram of the structure of a belt tear detection device based on three-dimensional vision provided in an embodiment of the present application; Figure 4 A schematic diagram of the structure of a computing device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0022] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0023] In some of the processes described in the specification and claims of this application and the above-mentioned figures, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this article or executed in parallel. The sequence numbers of the operations, such as 11, 12, etc., are only used to distinguish between different operations, and the sequence numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this article are used to distinguish different messages, devices, modules, etc., do not represent the order of precedence, and do not limit the "first" and "second" to different types.

[0024] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of this application.

[0025] In order to solve the problem of low belt tear detection accuracy caused by the difficulty in effectively distinguishing real cracks from surface interference, the inability to quantify crack depth and insufficient environmental robustness in the prior art, an embodiment of the present application provides a belt tear detection method based on three-dimensional vision. The method constructs a detection framework with three-dimensional point cloud data as the core, and adopts the following ideas: first, the belt surface image is collected by collaboratively using a line laser and an industrial camera, and then the three-dimensional point cloud data of the laser arc is calculated, breaking through the dependence of two-dimensional vision technology on planar features; based on the spatial mutation characteristics of the crack in the direction perpendicular to the tangent of the laser arc, a set of candidate points is screened to preliminarily eliminate planar interference; further combined with dynamic neighborhood breadth-first search and three-dimensional connected domain analysis, continuous crack morphology is extracted from the spatial topological structure dimension, and isolated noise is filtered through length and depth dual thresholds; finally, multi-cycle detection results are integrated to verify crack persistence, ensure the anti-interference and reliability of detection results in dynamic scenarios, and form a closed-loop detection logic from three-dimensional feature extraction to dynamic verification.

[0026] Figure 1 A flowchart of a belt tear detection method based on three-dimensional vision provided in an embodiment of the present application is shown in FIG. Figure 1 As shown, the method includes: S11. The belt surface image is collected by deploying a line laser and an industrial camera under the belt, and the three-dimensional point cloud data of the laser arc in the belt surface image is calculated.

[0027] Among them, the line laser is a device that emits a linear laser beam and is used to form measurable light stripes on the surface of an object. The industrial camera is a high-precision imaging device that is used to capture the laser arc image and transmit it to the processing system. Exemplarily, the industrial camera can be a high-speed industrial camera. The three-dimensional point cloud data is a set of discrete points composed of three-dimensional space coordinates, which characterizes the surface morphology of the belt. The three-dimensional point cloud data of the laser arc in the belt surface image can be calculated based on the baseline distance between the camera and the laser, the angle between the laser ray and the baseline, and the angle between the camera and the baseline, and the three-dimensional point cloud data is generated by the principle of optical triangulation.

[0028] In the embodiment of the present application, a laser beam is emitted by a line laser deployed under a high-speed conveyor belt to form a laser arc on the belt surface, and a high-speed industrial camera simultaneously captures the arc image at a preset angle (i.e., a fixed angle is formed between the high-speed industrial camera and the laser baseline). Based on the principle of optical triangulation, combined with the known baseline distance between the center of the camera and the center of the laser, the angle between the laser ray and the baseline, and the angle between the camera and the baseline, the three-dimensional coordinates of the laser arc in each frame of the image are calculated. The specific process is: through the geometric relationship between the laser plane equation and the camera projection model, the two-dimensional coordinates of each pixel point in the belt surface image are converted into three-dimensional space coordinates, and finally the three-dimensional point cloud data of the belt surface is generated.

[0029] S12, calculating the spatial distance value between each target point and the adjacent points in the three-dimensional point cloud data in the direction perpendicular to the tangent of the laser arc, and screening the points whose spatial distance value exceeds the preset distance threshold to form an initial candidate point set.

[0030] Among them, the laser arc tangent direction is the vector direction of the local geometric tangent of the laser arc, which is used to define the normal plane. The spatial distance value is the Euclidean distance between two points in the normal plane. Each target point in the three-dimensional point cloud data is also the current point. The initial candidate point set is formed by calculating and screening the spatial distance value between each target point and its adjacent points in the three-dimensional point cloud data in the direction perpendicular to the laser arc tangent.

[0031] In the embodiment of the present application, for each point in the three-dimensional point cloud data, its spatial feature (X vector) is calculated, and the spatial feature is used to reflect the spatial distance difference between the point and its surrounding neighboring points in the direction perpendicular to the tangent of the laser arc. The specific process is: first, the local tangent direction of the laser arc is fitted by the difference in the coordinates of the adjacent points, and the normal plane perpendicular to the tangent direction is determined; then the Euclidean distance between the target point and the neighboring point in the normal plane is calculated. If the distance difference exceeds a preset threshold (such as 1mm), the point is marked as a candidate point. All candidate points that meet the conditions constitute a set A (initial candidate point set), where set A may contain crack points or noise points.

[0032] S13. Starting from each point in the initial candidate point set, the points that are adjacent in space and belong to the initial candidate point set are expanded through a breadth-first search algorithm to obtain a target candidate point set, so as to form multiple disconnected connected domains, and calculate the length and depth of each connected domain.

[0033] Among them, the breadth-first search algorithm is a graph traversal algorithm that expands adjacent nodes hierarchically to find connected areas. A connected domain refers to an independent area composed of interconnected points in space. The target candidate point set is a connected area point set obtained by expanding the initial candidate point set through the breadth-first search algorithm. For example, the formula for calculating the length of the connected domain is: ,in, L is the length of the connected domain, are the maximum and minimum x-axis coordinates of all points in the connected domain projected on the belt surface, is the maximum and minimum y-axis coordinates of all points in the connected domain projected on the belt surface. For example, the formula for calculating the depth of the connected domain is: ,in, D is the depth of the connected domain, is the height value of each point in the connected domain, It is the reference height of the belt surface.

[0034] In the embodiment of the present application, for each point in the initial candidate point set, a breadth-first search algorithm is used to expand the connected domain: starting from the current point, check whether its spatial adjacent points belong to the initial candidate point set. If they do, they are included in the same connected domain, and the adjacent points of the adjacent points are recursively searched until no new candidate points can be found. All disconnected connected domains constitute the target candidate point set, and each connected domain represents an independent suspected crack area.

[0035] S14, filtering out the connected domains that do not meet the threshold condition among all the connected domains, retaining the connected domains that meet the threshold condition, treating all the connected domains that meet the threshold condition as corresponding potential cracks, and forming a potential crack set with the spatial location information of all the potential cracks. The domains that do not meet the threshold condition are those whose length and depth do not reach the corresponding preset threshold.

[0036] Among them, the length threshold usually refers to the main axis length or contour perimeter of the connected domain. Cracks usually appear in a long and thin shape, and a too short connected domain may be noise rather than a real crack. The depth threshold refers to the gray value difference or the actual three-dimensional depth information. A connected domain with insufficient depth may indicate low contrast or shallow cracks, which are difficult to be determined as valid cracks. Only when the length and depth of the connected domain do not reach the threshold at the same time, it is judged as "not meeting the conditions". This means that the retention condition is length ≥ length threshold or depth ≥ depth threshold, and either one of them can be met. The filtering condition is length < length threshold and depth < depth threshold. When both are not met at the same time, it means that the threshold condition is not met. The spatial position of the potential crack is the coordinate set of the detected crack connected domain in the image or three-dimensional space.

[0037] In the embodiment of the present application, each connected domain in the target candidate point set is subjected to threshold screening, and the screening process may be: calculating the length (i.e., the straight-line distance through its two farthest ends) and depth of the connected domain. If both the length and the depth do not reach the preset threshold, it is determined to be noise and removed; otherwise, it is retained as a potential crack, and its spatial position information is stored in the potential crack set.

[0038] S15. During the belt circulation process, if there is a real crack, a belt tear detection result including the spatial position information of the real crack and the corresponding alarm level is generated. The real crack is a potential crack recorded in the same spatial position in the potential crack set in multiple consecutive operation cycles.

[0039] Among them, the real cracks are potential cracks that recur across cycles and have been verified to be consistent in time and space. The alarm level is the severity of the alarm divided according to the depth of the crack (such as low danger, medium danger and high danger). The final generated real crack spatial location information can be in the following two forms: merging the three-dimensional point cloud data of the crack detected in all cycles to form a spatiotemporal fusion point cloud to reflect the complete morphology and expansion trend of the crack. Taking the weighted average of the center coordinates of the crack in all cycles to ensure that the position information reflects the spatial position of the crack in real time.

[0040] In the embodiment of the present invention, combined with the cyclic operation characteristics of the conveyor belt, the potential crack set is verified across cycles. If a crack at a certain spatial position is detected in multiple consecutive operation cycles, it is determined to be a real crack. The alarm level is generated according to the crack depth, and the belt tear detection result including the position, size and alarm level is output.

[0041] Here is a specific example: In the steel belt transportation scenario, the line laser projects a green laser line onto the running belt, and the industrial camera captures images at a rate of 30 frames per second. The system calculates the three-dimensional point cloud data in real time and finds that the distance of a certain point cloud in the normal plane suddenly changes to 3mm, marking it as the initial candidate point. The point is expanded through the breadth-first search algorithm to form a connected domain with a length of 15cm and a depth of 4mm. After filtering, the connected domain is retained as a potential crack because it exceeds the threshold. In the subsequent three belt cycles, the system detected the crack at the same coordinates, and finally triggered a high-risk alarm, prompting a shutdown for maintenance.

[0042] By executing steps S11 to S15, the embodiment of the present application improves the accuracy and robustness of belt tear detection through three-dimensional point cloud analysis combined with multi-cycle verification: the line laser and the camera work together to achieve high-resolution morphology reconstruction; through spatial distance screening and connected domain extraction, it can effectively distinguish between real cracks and instantaneous noise; the cross-cycle matching mechanism further reduces the false alarm rate and ensures the credibility of the alarm results.

[0043] Specifically, this application is based on the three-dimensional point cloud data of laser arcs collected by line lasers and industrial cameras, which breaks through the dependence of traditional two-dimensional visual technology on plane information, can accurately characterize the depth changes of belt surface morphology, and improve the detection sensitivity of three-dimensional spatial features of tiny cracks (especially longitudinal tearing). By calculating the spatial distance between adjacent points in the three-dimensional point cloud in the direction of the vertical laser arc tangent, and combining the preset distance threshold to screen candidate points, it can effectively distinguish between real cracks and surface stains, reflections and other plane interferences, reduce the false detection rate caused by environmental noise, and improve the accuracy of belt tear detection results. The breadth-first search algorithm is used to generate interconnected domains that are not connected to each other, and combined with the length and depth double threshold filtering, the pseudo-crack areas formed by isolated noise points are accurately eliminated; at the same time, the potential cracks at the same spatial position are verified through multi-cycle continuous detection to avoid false alarms caused by instantaneous interference, and ensure the reliability and stability of the detection results. By generating crack spatial position and alarm level information in real time, accurate decision support is provided for belt operation and maintenance, meeting the needs of unmanned and high-efficiency tear risk warning in industrial scenarios.

[0044] In a possible embodiment, S13, starting from each point in the initial candidate point set, expanding points that are spatially adjacent and belong to the initial candidate point set by a breadth-first search algorithm to obtain a target candidate point set, including: Step 131, traverse each target point in the initial candidate point set, mark the target point as a member of the current connected domain, and insert it into the preset queue to be viewed, and calculate the dynamic neighborhood range of the current target point according to the belt running direction and belt running speed.

[0045] Among them, the search distance extended along the belt running direction is greater than the direction perpendicular to the tangent of the laser arc. The extended distance of the dynamic neighborhood range in the belt running direction is: L = v ×Δ t + L 0, where v is the real-time belt speed, Δ t is the image acquisition cycle, L 0 is the base neighborhood length.

[0046] In the embodiment of the present application, the target point is marked as a member of the current connected domain by a designated field corresponding to the current connected domain. For example, the marking information of the target point A1 is the designated field LTY1, indicating that the target point A1 is a member of the connected domain 1; the marking information of the target point A2 is the designated field LTY2, indicating that the target point A2 is a member of the connected domain 2. The extension length of the dynamic neighborhood in the direction of belt operation is the speed multiplied by the time window, and the horizontal width is fixed to a preset value to form a dynamic rectangular search area. This step ensures the spatiotemporal consistency of the neighborhood search by dynamically adjusting the neighborhood range to adapt to the spatial displacement of the candidate points caused by the belt movement.

[0047] Step 132: Search for adjacent points in the dynamic neighborhood along a direction perpendicular to the tangent of the laser arc that are spatially adjacent and belong to the initial candidate point set.

[0048] Among them, the judgment condition of spatial proximity is that the three-dimensional Euclidean distance between the adjacent point and the current point is less than the preset connectivity distance threshold, and the three-dimensional distance between the adjacent point and the current point in the dynamic neighborhood is less than the resolution threshold. The initial candidate point set is a set of points whose spatial distance exceeds the threshold, including potential crack points and noise points. The belt running direction is the direction in which the belt transmits the material, usually measured by an encoder or speed sensor. The tangent direction of the laser arc is the local tangent direction of the laser line on the belt surface, which is calculated by the difference of the coordinates of adjacent points.

[0049] In the embodiment of the present invention, the specific process of step 132 is: taking the current target point as the center, dividing the dynamic neighborhood into three-dimensional grids, traversing all grid points, and if there is an initial candidate point in the three-dimensional point cloud data corresponding to a grid point, it is determined to be a spatial adjacent point. This step focuses on the deformation characteristics in the vertical direction of the crack by constraining the search direction and range, reducing interference in non-related directions.

[0050] Step 133: Determine whether the neighboring points of the target point belong to the initial candidate point set and are not marked.

[0051] In the embodiment of the present invention, the specific implementation of step 133 is: check whether the unique identifier (such as the coordinate hash value) of the adjacent point exists in the hash table of the initial candidate point set, and query the mark status table to confirm that it is not marked as a member of any connected domain. If the condition is met, go to step 134; otherwise, skip the adjacent point and continue to search for other neighboring points.

[0052] Step 134, if yes, mark the adjacent point as a member of the current connected domain, insert it to the end of the preset queue to be viewed, take the adjacent point as the new target point, repeat the marking, inserting, calculating, searching, and judging steps until all the target points in the preset queue to be viewed form a connected domain, clear the preset queue to be viewed, and repeat the process of forming a new connected domain until each target point in the initial candidate point set is marked.

[0053] The preset queue to be checked is a first-in-first-out data structure for temporarily storing the connected domain points to be expanded. The marking state table is a hash table that records whether each point has been assigned to a connected domain.

[0054] In the embodiment of the present invention, if the adjacent point meets the conditions, it is marked as a member of the current connected domain, inserted into the end of the preset queue to be checked, and the adjacent point is used as the new target point. Until the preset queue to be checked is empty, it means that the current connected domain expansion is completed. Then the queue is cleared, and the next unmarked point is selected from the initial candidate point set, and the above process is repeated until all points are marked. This process ensures the integrity of the connected domain through iterative expansion.

[0055] Step 135: Delete all target points in the initial candidate point set that are not in any connected domain to obtain a target candidate point set.

[0056] Among them, the target candidate point set is the candidate point set retained after the connected domain is expanded, excluding isolated noise points.

[0057] In the embodiment of the present invention, the specific operation of step 135 is: traverse the initial candidate point set, and if the mark status of a point is "unowned", remove it from the set. The points finally retained constitute the target candidate point set, and each connected domain thereof represents a suspected crack area for subsequent threshold filtering and cross-cycle verification.

[0058] Here is a specific example: In the iron ore belt transportation system, when the belt runs at a speed of 2m / s, the system detects the initial candidate point. According to the belt running speed of 2m / s and the time window of 0.1s, the dynamic neighborhood range of the point is calculated as [4.8m, 5.2m] in the x direction and [0.15m, 0.25m] in the y direction. It is marked as a member of the current connected domain and inserted into the queue to be viewed. It searches in the dynamic neighborhood along the normal plane direction (y axis) and finds 3 adjacent points belonging to the initial candidate point set. It verifies that all 3 points are not marked, enters the expansion process, marks the 3 points and inserts them to the end of the queue, recursively expands with the new point as the target, and finally forms a connected domain containing 12 points. After clearing the queue, it traverses the remaining unmarked points, deletes 20 unmarked isolated points in the initial candidate point set, and retains 3 connected domains to form the target candidate point set. In the subsequent processing, one connected domain triggers a high-risk alarm due to exceeding the threshold, and the rest are filtered. The system accurately locates the crack position and triggers shutdown and maintenance to avoid belt tearing accidents.

[0059] By executing steps 131 to 135, the embodiment of the present application improves the robustness and accuracy of crack detection through a dynamic neighborhood range and a directional search strategy, adapts to the displacement of candidate points caused by belt movement, avoids missed detection due to speed changes, focuses on the vertical deformation characteristics of the cracks, suppresses the lateral texture interference of the belt surface, ensures the integrity of the connected domain, and efficiently filters instantaneous noise, which is suitable for industrial scenarios with different belt speeds and crack morphologies.

[0060] In a possible embodiment, in step 132, the judgment condition of spatial adjacency is that the three-dimensional Euclidean distance between the adjacent point and the target point is less than a preset connectivity distance threshold, and the preset connectivity distance threshold is dynamically set according to the mapping relationship between the spot diameter of the line laser calibrated experimentally and the belt running speed.

[0061] The adjacent points in the search space that are adjacent and belong to the initial candidate point set include: Step a1, calculate the tangent direction vector of the laser arc at the target point.

[0062] The laser arc tangent direction vector is the unit vector of the local geometric tangent of the laser arc, which represents the tendency of the crack to expand along the running direction of the belt.

[0063] Step a2: Generate a vector perpendicular to the tangent direction of the laser arc based on the tangent direction vector of the laser arc and the unit vector of the belt thickness direction.

[0064] The belt thickness direction unit vector is a preset direction vector perpendicular to the belt surface and is used to define the belt thickness direction.

[0065] Step a3: Extend the preset connected distance threshold along the positive and negative directions of the tangent direction vector of the laser arc to form a strip-shaped search area, and determine the points in the strip-shaped search area except the target point as adjacent points.

[0066] The preset connection distance threshold is a preset length value (such as 10 cm) for extending the search range along the tangent direction, which controls the maximum extension distance of the connected domain. The strip search area is a narrow three-dimensional area with the tangent direction as the main axis and the normal plane direction as the secondary axis, which is used for directional search of adjacent points.

[0067] Here is a specific example: In the steel belt detection, the system detects the target point, takes the three adjacent points before and after the point, and obtains the tangent direction vector by least squares fitting. =(0.98,0.02,0), cross product and the belt thickness direction vector (0,0,1)(0,0,1) to generate the normal plane direction vector =(-0.02,0.98,0), extending 10cm in the positive and negative directions along the tangent, with a width of ±2mm in the normal plane direction, constructing a strip-shaped area (x range [2.9m,3.1m], y range [0.498m,0.502m]), selecting 5 adjacent points, and expanding through the connected domain to form a crack with a length of 12cm and a depth of 3mm. After cross-cycle verification, the alarm is triggered to guide maintenance personnel to accurately locate and repair.

[0068] By executing steps a1 to a3, the embodiment of the present application constrains the search direction, reduces interference from non-related areas, and improves crack detection efficiency. The strip-shaped area fits the characteristics of cracks extending along the belt running direction to avoid missing oblique or curved cracks. The preset connection distance threshold allows the detection sensitivity to be dynamically adjusted according to actual working conditions (such as belt speed, crack length).

[0069] Exemplarily, the present application embodiment provides a specific embodiment of a belt tear detection method based on three-dimensional vision, as described below: like Figure 2 As shown, the laser beam emitted by the line laser irradiates the high-speed conveyor belt to form an arc. The camera captures the arc at a certain angle, and uses an algorithm to perform corresponding processing on each frame of the belt surface image. Under the conditions of the known baseline distance between the center of the camera and the center of the line laser, the angle between the laser beam and the baseline, and the angle between the camera and the baseline, the spatial three-dimensional coordinates of the arc can be calculated according to the principle of optical triangulation. The spatial three-dimensional coordinates of the arc can be represented in the form of point cloud data. If there are cracks on the surface of the belt, there will be uneven ups and downs on the point cloud, but because the cracks in the belt may be very small, such as about 1mm wide, the point cloud has very small ups and downs. In order to detect such tiny ups and downs without frequent misdetections, the embodiment of the present application proposes a crack detection method based on point cloud connected domain calculation. The method can execute the following process: First, calculate the spatial feature (X vector) of each point on the point cloud. The spatial feature is the spatial distance difference between each point and its surrounding points in the vertical direction (i.e., perpendicular to the arc tangent direction). Find the points whose distance difference (X vector) is greater than a certain threshold, and form these points into a set A. These points may be crack points or noise points. Noise points are generally discontinuous, while crack points may generally be continuous points.

[0070] Secondly, for each point in set A, calculate its surrounding connected domain. That is, starting from each point in A, check whether its spatially adjacent points belong to A. If so, assign the adjacent points to the same connected domain of the point, and then continue to expand the search from the adjacent points. If the adjacent points of the adjacent points also belong to A, they are also assigned to the same connected domain of the point, and then continue to expand the search from the adjacent points of the adjacent points until the spatially adjacent points encountered do not belong to A, and the expansion of the connected domain stops. All disconnected connected domains constitute set B.

[0071] Again, the size of each connected domain in set B is calculated. If the length and depth are greater than a certain threshold, it is considered that this connected domain may be a crack, and its spatial location information is recorded. These spatial location information constitute set C.

[0072] Finally, since the conveyor belt rotates in a circular motion, if crack information is periodically recorded at a certain spatial position (that is, the position already exists in set C), it is considered that the belt is torn and an alarm prompt is output.

[0073] In this experiment, the hardware includes a line laser and an industrial camera, both of which are placed under the belt. The line laser is 500mm away from the camera, the line laser's ray direction is at an angle of about 90 degrees to the baseline, the camera is at an angle of about 45 degrees to the baseline, the belt is about 1000mm wide, the distance between the belt and the laser is about 400mm, and there are several cracks with different widths of 1mm-10mm on the belt.

[0074] When the crack of the belt runs to the position illuminated by the laser line, the image can be seen to be disconnected or fluctuating, and the point cloud display jumps, indicating that the point cloud has a large difference in the vertical distance in space. At this time, the system outputs the spatial position information of the crack. The belt runs in a cycle. If there is a crack record at the same position every time, the system will alarm. The experimental results show that the crack detection accuracy can reach the millimeter level.

[0075] The specific steps of connected domain calculation are: Step 1: Starting from any point x that meets the detection feature attribute (belonging to A), mark x as belonging to a connected domain S. Create a queue Q to be checked, insert this point x into the queue Q, and currently the queue Q has only one element x.

[0076] Step 2: Take an element from the head of queue Q (starting with x), set it as the current point, and look at the features of the spatial left and right adjacent points of the current point. If the adjacent point has not been viewed and the feature is the same as the feature attribute of the current point (belonging to A), then this point is marked as belonging to the connected domain S and inserted to the end of the queue to be viewed Q. The current point is marked as having been viewed and deleted from the Q queue.

[0077] Step 3: Repeat step 2 until the queue Q is empty. Then the connected domain S is found.

[0078] Step 4: Repeat the steps from 1 until all point cloud points have been viewed.

[0079] Finally, for each connected domain, the spatial shape characteristic parameters of the connected domain, including length, depth, etc., are calculated. If a certain threshold condition is met, it is classified as a crack.

[0080] The embodiment of the present application is based on computer three-dimensional vision technology to achieve automatic detection of longitudinal tears in the conveyor belt, replacing manual detection, avoiding potential personnel safety accidents, timely identification, timely alarm, timely stopping of the belt, and timely repair of any damage, avoiding safety hazards, and greatly improving the safe operation level of the belt.

[0081] In a possible embodiment, S12, calculating the spatial distance value between each target point and the adjacent point in the three-dimensional point cloud data in the direction perpendicular to the tangent of the laser arc, screening the points whose spatial distance value exceeds the preset distance threshold, and forming an initial candidate point set, including: Step b1: obtaining cross-sectional data of the multi-layer composite structure through a preset belt cross-sectional scanning device, and establishing a correlation model between belt material and belt thickness.

[0082] Among them, the belt cross-section scanning device is a detection device used to obtain the internal parameters of the multi-layer structure of the belt. The multi-layer composite structure can include a rubber cover layer, a fiber reinforcement layer and a wire rope core layer. The correlation model is a mathematical model that describes the relationship between the belt material properties and thickness, which is established through regression analysis.

[0083] Step b2: dynamically generate a hierarchical distance threshold for the hierarchical structure where each target point is located according to the association model.

[0084] Among them, the layer distance threshold is the depth difference tolerance value that is dynamically adjusted according to the material layer characteristics and is used to filter outliers.

[0085] Step b3: Calculate the spatial distance difference between the target point and the adjacent points in the anisotropic neighborhood in the belt thickness direction. Compare the spatial distance difference with the layered distance threshold, filter out the abnormal points whose spatial distance difference exceeds the layered distance threshold, and determine the maximum penetration depth and layered position information of each abnormal point.

[0086] Among them, the anisotropic neighborhood is a neighborhood with different extension ranges defined in the X-axis, Y-axis and Z-axis directions to adapt to the layered structure characteristics of the belt. The maximum penetration depth is the maximum coordinate difference between the abnormal point and the target point in the thickness direction, reflecting the severity of the deformation.

[0087] Step b4: construct a density field based on the spatial distribution characteristics of the outliers, and calculate the spatial concentration of each outlier in the density field within the preset volume element.

[0088] The density field is a three-dimensional grid data field that characterizes the spatial distribution density of outliers. The spatial aggregation degree is the ratio of the number of outliers in a unit volume, which is used to quantify the aggregation characteristics. The size of the volume element is proportional to the anisotropic neighborhood search range.

[0089] Step b5: determine the product of the spatial concentration and the maximum penetration depth as a comprehensive anomaly score, and screen outliers whose comprehensive anomaly scores exceed a preset score threshold to form an initial candidate point set.

[0090] Among them, the comprehensive anomaly score is a weighted indicator that integrates aggregation and depth to improve the reliability of crack judgment.

[0091] Here is a specific example: In the steel transport belt inspection, the laser tomography scanner obtains the belt cross-section data, establishes the correlation model between the rubber layer (thickness 8mm) and the fiber layer (thickness 5mm), and the delamination distance threshold of the fiber layer is set as =1.2mm, 3 abnormal points are found in the neighborhood of the target point, the maximum penetration depth is 2.0mm, exceeding the threshold of 1.2mm, the abnormal points are concentrated in the 1cm³ volume element, and the spatial aggregation degree =0.85. The comprehensive score S=0.85×2.0=1.7 (exceeding the threshold of 0.7) is included in the initial candidate point set. In the subsequent steps, the area is confirmed as a real crack and an alarm is triggered after the connected domain expansion and cross-cycle verification to avoid delamination accidents.

[0092] By executing steps b1 to b5, the embodiment of the present application avoids misjudgment due to material differences through dynamic adaptation of multi-layer materials and fusion of multiple features, adapts the multi-layer structure of the composite belt, focuses on deformation in the thickness direction, suppresses plane texture interference, distinguishes random noise from continuous cracks through spatial aggregation features, and integrates physical depth and distribution characteristics to reduce the missed detection rate.

[0093] In a possible embodiment, step b3, determining the maximum penetration depth and layered position information of each abnormal point, includes: Step c1: According to the spatial position information of the abnormal point and the geometric model of the layer boundary, the maximum deviation between each abnormal point and the layer boundary is calculated to generate the maximum penetration depth.

[0094] Among them, anomalies are discrete points in the detection data that deviate from the normal range, which may indicate defects such as cracks and holes. Spatial position information is the coordinate data of the anomaly. The layer boundary is the geometric boundary that divides different material layers or structural layers. Geometric data accurately describes the spatial position, shape and range of each layer through, for example, plane equations, surface parameter equations, point clouds or three-dimensional mesh models. In this application, the geometric model of the layer boundary is the core basis for calculating the maximum deviation of the anomaly. The maximum deviation is the maximum vertical distance of the anomaly relative to the layer boundary, which distinguishes internal and external penetration.

[0095] Step c2: Combine the hierarchical index table to perform hierarchical position analysis on the spatial position information of the abnormal point to obtain hierarchical position information.

[0096] The hierarchical index table is a mapping table that records the hierarchical number, spatial range and attributes. Hierarchical location analysis is the spatial matching and attribute analysis process of associating anomalies to specific hierarchies. Hierarchical location information is structured data that includes hierarchical number, penetration status and depth.

[0097] The following is a specific example: The conveyor belt of a steel plant is composed of three layers of composite materials, among which the outer layer (layer 1): wear-resistant rubber layer, thickness 5mm (z range: 0mm-5mm). Middle layer (layer 2): steel wire reinforcement layer, thickness 5mm (z range: 5mm-10mm). Inner layer (layer 3): tear-resistant fiber layer, thickness 5mm (z range: 10mm-15mm). Through the three-dimensional visual inspection system, the coordinates of a certain abnormal point are found to be (x=200mm, y=50mm, z=7mm), and its maximum penetration depth and layer position information need to be determined. The lower boundary plane equation of the outer rubber layer: z=5 mm, the lower boundary plane equation of the middle steel wire layer: z=10 mm, the abnormal point z=7 mm, located in layer 2 (steel wire reinforcement layer), the vertical distance relative to the lower boundary of the outer layer (z=5mm) is 7-5=2mm. The vertical distance relative to the lower boundary of the middle layer (z=10mm) is 10-7=3mm. The penetration depth from the outer layer to the middle layer is 2mm, and the layers are steel wire reinforcement layers. The rubber layer (2mm) has been penetrated, but the steel wire layer has not been completely penetrated, and the tear-resistant fiber layer has not been touched. If only relying on two-dimensional images, the abnormal point z=7mm may be misjudged as a surface scratch (the depth cannot be distinguished), resulting in missed detection of steel wire layer damage. Three-dimensional detection combined with a layered model can identify internal structural risks and improve detection reliability.

[0098] By executing steps c1 to c2, the embodiment of the present application realizes accurate spatial positioning and hierarchical management of abnormal points through geometric penetration depth calculation and hierarchical attribution analysis. The degree of damage to the boundary by the abnormal points is quantified to avoid subjective misjudgment; combined with the hierarchical index table, the abnormal points are dynamically associated with the hierarchical attributes of the engineering structure, supporting applications such as crack expansion trend analysis and safety hazard classification, and improving the interpretability of detection data and decision-making efficiency.

[0099] In a possible embodiment, S14, spatial location information of all potential cracks is used to form a potential crack set, including: Step 141: extract crack feature points according to the spatial positions of all potential cracks, and form an initial potential crack point set.

[0100] Among them, the crack feature points are points with geometric changes in the crack contour. The initial potential crack point set is a discrete point set composed of feature points, which characterizes the rough spatial distribution of the cracks.

[0101] Step 142: spatially cluster the initial potential crack point set and merge adjacent points to form potential crack segments.

[0102] Among them, spatial clustering is an algorithm that merges neighboring points into the same group based on distance measurement. Merging neighboring points determines the maximum Euclidean distance threshold for determining whether two points belong to the same cluster. Potential crack segments are continuous curve segments generated by clustering and fitting, representing local crack morphology.

[0103] Step 143: Calculate the geometric features of all potential fracture segments, analyze the adjacency and intersection relationships between fracture segments, and construct a fracture topology map.

[0104] Among them, geometric features are parameters that describe fracture segments. Adjacency relations are spatial connections between fracture segments with very close endpoints. Crossover relations are relationships where fracture segments intersect or overlap in a two-dimensional plane or three-dimensional space. Fracture topology is a graph structure model with fracture segments as nodes and connection relations as edges.

[0105] Step 144: Perform global optimization based on geometric continuity and the fracture topology map, remove low-confidence fracture segments, and obtain a potential fracture set.

[0106] Among them, geometric continuity refers to the smooth transition characteristics of the fracture segment in terms of direction, curvature, etc. Low confidence fracture segment: a suspected noise segment that does not meet the continuity or topological connection conditions.

[0107] The following is a specific example: In the 3D visual inspection of the conveyor belt of a steel plant, the 3D point cloud data of the belt surface is collected by line laser and industrial camera, and 52 initial potential crack feature points distributed in the middle of the conveyor belt are detected. These points are irregularly distributed and the local curvature changes significantly; then, the system uses a spatial clustering algorithm based on Euclidean distance (threshold is 3mm) to merge the 52 feature points into 6 continuous crack segments. For example, one of the crack segments consists of 15 feature points, extending 28cm along the running direction of the belt, with an average curvature of 0.05 / mm; then, the system analyzes the geometric features and spatial relationship of each crack segment. When it is found that segment 1 (length 32cm) and segment 2 (length 18cm) are at the coordinates (x=1.5m, When the endpoint spacing at y=0.3m) is only 1.2mm, it is determined to be an adjacent relationship, and segment 3 (length 12cm) and segment 4 (length 9cm) cross in three-dimensional space to form an "X"-shaped structure, and a topological graph is constructed with crack segments as nodes and connection relationships as edges; finally, based on the principle of geometric continuity (directional deviation <5°, curvature change rate <15%), the system eliminates segment 5 (curvature mutation 27%) and segment 6 (isolated and unconnected), and retains 4 high-confidence crack segments (total length 71cm) to form the optimized potential crack set, and triggers a secondary alarm to guide maintenance personnel to focus on checking the steel wire reinforcement layer in the 1.2m to 1.9m area of ​​the middle section of the conveyor belt to avoid production accidents caused by longitudinal tearing expansion.

[0108] By executing steps 141 to 144, the embodiment of the present application realizes separation of real cracks from discrete noise, extracts feature points, converts cluttered connected domains into discrete point sets, generates continuous crack segments through clustering, enhances morphological consistency, reveals crack connection rules by quantifying geometric characteristics and constructing topological relationships, integrates physical laws and topological logic, removes abnormal segments, and finally outputs a high-precision potential crack set. This process improves the accuracy, continuity and interpretability of crack detection and is suitable for health monitoring of complex engineering structures such as bridges and tunnels.

[0109] In a possible embodiment, S13, calculating the length and depth of each connected domain, includes: Step d1: parse the spatial positions and boundary information of all connected domains to generate an initial description information set for each connected domain.

[0110] Among them, the spatial position is the position parameter of the connected domain in the image coordinate system. The boundary information is the geometric characteristics of the connected domain outline. The initial description information set is a structured data set containing the basic attributes of the connected domain.

[0111] Step d2: Based on the initial description information set and the preset spatial geometric relationship model, perform boundary extraction for each connected domain to obtain the coordinates of all boundary points of each connected domain.

[0112] The preset spatial geometric relationship model is an algorithm or rule for describing geometric relationships. The boundary point coordinates are pixel coordinate sequences of the connected domain contour arranged in order.

[0113] Step d3: Analyze the relative position relationship between all boundary points of each connected domain according to the coordinates of all boundary points of each connected domain, determine the two first endpoints corresponding to the longest path, and use the distance value between the two first endpoints as the length of the connected domain.

[0114] The relative position relationship is the geometric relationship between the boundary points. The longest path is the maximum Euclidean distance line segment among the boundary points of the connected domain. The first endpoints are the two endpoints of the longest path. The length of the connected domain is the Euclidean distance between the first endpoints.

[0115] Step d4: Analyze the distance distribution between all boundary points of each connected domain and the preset connected domain center point according to the coordinates of all boundary points of the connected domain, determine the two second endpoints perpendicular to the longest path, and use the distance value between the two second endpoints as the depth of the connected domain.

[0116] Among them, the distance distribution is the statistical distribution of the projection distance from the boundary point to the center point or axis. The center point of the connected domain is the arithmetic average position of the boundary point coordinates. The second endpoint is the maximum distance point pair perpendicular to the longest path direction. The depth of the connected domain refers to the Euclidean distance between the second endpoints.

[0117] The following is a specific example: In the 3D visual inspection of the conveyor belt in a steel plant, the system identifies a connected domain on the belt surface. First, its spatial position is parsed as a rectangular area in the image coordinate system (upper left corner coordinates (150, 80), lower right corner coordinates (320, 120)). The boundary contour contains 58 discrete points, and an initial description information set is generated, including area, perimeter, and circumscribed rectangle parameters. Then, the boundary point coordinate sequence is extracted based on the Moore neighborhood tracking algorithm, and is arranged clockwise as [(150,80),(155,82), ..., (320,120)]; then traverse all boundary point pairs to calculate the Euclidean distance, determine that the longest path endpoints are (160,85) and (310,115), and the straight-line distance between the two is 152mm, which is marked as the length of the connected domain; further calculate the arithmetic mean of the boundary points to get the center point (235,100), establish a vertical axis along the longest path direction, select the two second endpoints (230,70) and (230,130) farthest from the center point, and measure the depth of 60mm; combined with the length and depth data (152mm×60mm), the system determines that the connected domain meets the characteristics of a longitudinal crack, triggers a first-level alarm and outputs three-dimensional coordinates (1.5m, 0.6m), and guides maintenance personnel to give priority to checking the junction of the rubber covering layer and the steel wire layer in the roller area of ​​the third section of the conveyor belt to avoid belt breakage and shutdown due to deep tearing.

[0118] By executing steps d1 to d4, the embodiment of the present application first analyzes the spatial position and boundary information of all connected domains, generates an initial description information set containing parameters such as the coordinates, area, and perimeter of the circumscribed rectangle; then extracts the boundary point coordinates of each connected domain based on the initial description information set and the preset spatial geometric relationship model to form a set of contour points arranged in order; then analyzes the relative position relationship between the boundary points, calculates the Euclidean distance between all point pairs, selects the two first endpoints with the largest spacing, and determines the length of the connected domain; finally, based on the center point of the connected domain and the direction of the longest path, analyzes the distribution of the projection distance of the boundary points, determines the two second endpoints perpendicular to the longest path, and calculates the depth value. This process converts the image connected domain into quantifiable length and depth parameters through layer-by-layer data conversion and geometric calculation, avoids human intervention errors, and is suitable for accurate detection and safety assessment of crack sizes in engineering structures such as bridges and tunnels.

[0119] Figure 3 A schematic diagram of the structure of a belt tear detection device based on three-dimensional vision provided in an embodiment of the present application is shown in FIG. Figure 3 As shown, the device comprises: The acquisition module 31 is used to acquire the belt surface image through a line laser and an industrial camera deployed under the belt, and calculate the three-dimensional point cloud data of the laser arc in the belt surface image.

[0120] The screening module 32 is used to calculate the spatial distance value between each target point and the adjacent points in the three-dimensional point cloud data in the direction perpendicular to the tangent of the laser arc, and screen the points whose spatial distance value exceeds the preset distance threshold to form an initial candidate point set.

[0121] The calculation module 33 is used to start from each point in the initial candidate point set, expand the points that are adjacent in space and belong to the initial candidate point set through a breadth-first search algorithm, obtain a target candidate point set, so as to form multiple disconnected connected domains, and calculate the length and depth of each connected domain.

[0122] The filtering module 34 is used to filter the connected domains that do not meet the threshold conditions in all connected domains, retain the connected domains that meet the threshold conditions, regard all the connected domains that meet the threshold conditions as corresponding potential cracks, and form a potential crack set with the spatial position information of all potential cracks. The domains that do not meet the threshold conditions are those whose length and depth do not reach the corresponding preset thresholds.

[0123] The generation module 35 is used to generate a belt tear detection result including the spatial position information of the real crack and the corresponding alarm level if there is a real crack during the belt circulation operation. The real crack is a potential crack recorded in the same spatial position in the potential crack set in multiple consecutive operation cycles.

[0124] Figure 3 The belt tear detection device based on three-dimensional vision can be performed Figure 1 The implementation principle and technical effect of the belt tear detection method based on three-dimensional vision described in the embodiment shown are not repeated here. The specific way in which each module and unit performs operations in the belt tear detection device based on three-dimensional vision in the above embodiment has been described in detail in the embodiment of the method, and will not be elaborated here.

[0125] In one possible design, Figure 3 A belt tear detection device based on three-dimensional vision in the embodiment shown can be implemented as a computing device, such as Figure 4 As shown, the computing device may include a storage component 41 and a processing component 42 .

[0126] The storage component 41 stores one or more computer instructions, wherein the one or more computer instructions are called and executed by the processing component 42 .

[0127] The processing component 42 is used to collect the belt surface image through the line laser and industrial camera deployed under the belt, and calculate the three-dimensional point cloud data of the laser arc in the belt surface image. Calculate the spatial distance value between each target point and the adjacent point in the three-dimensional point cloud data in the direction perpendicular to the tangent of the laser arc, and select the points whose spatial distance value exceeds the preset distance threshold to form an initial candidate point set. Starting from each point in the initial candidate point set, the points that are spatially adjacent and belong to the initial candidate point set are expanded through the breadth-first search algorithm to obtain the target candidate point set to form multiple interconnected domains, and calculate the length and depth of each connected domain. Filter the connected domains that do not meet the threshold conditions in all connected domains, retain the connected domains that meet the threshold conditions, and take all the connected domains that meet the threshold conditions as corresponding potential cracks. The spatial position information of all potential cracks constitutes a potential crack set. The threshold conditions are not met when the length and depth do not reach the corresponding preset threshold. During the belt cycle operation, if there is a real crack, a belt tear detection result containing the spatial position information of the real crack and the corresponding alarm level is generated. The real crack is a potential crack that is recorded in the same spatial position in the potential crack set in multiple consecutive operation cycles.

[0128] The processing component 42 may include one or more processors to execute computer instructions to complete all or part of the steps in the above method. Of course, the processing component may also be implemented by one or more application-specific integrated circuits (ASIC), digital signal processors (DSP), digital signal processing devices (DSPD), programmable logic devices (PLD), field programmable gate arrays (FPGA), controllers, microcontrollers, microprocessors or other electronic components to perform the above method.

[0129] The storage component 41 is configured to store various types of data to support operations at the terminal. The storage component can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as random access memory (RAM), static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0130] Of course, the computing device may also include other components, such as input / output interface, display component, communication component, etc. The input / output interface provides an interface between the processing component and the peripheral interface module, which may be an output device, an input device, etc.

[0131] The communication component is configured to facilitate, among other things, wired or wireless communications between the computing device and other devices.

[0132] Among them, the computing device can be a physical device or an elastic computing host provided by a cloud computing platform, etc. In this case, the computing device can refer to a cloud server, and the above-mentioned processing components, storage components, etc. can be basic server resources rented or purchased from the cloud computing platform.

[0133] The present application also provides a computer storage medium storing a computer program, wherein the computer program can achieve the above-mentioned Figure 1 The illustrated embodiment is a belt tear detection method based on three-dimensional vision.

[0134] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0135] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0136] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0137] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A belt tear detection method based on three-dimensional vision, characterized in that: include: The belt surface image is collected by a line laser and an industrial camera deployed under the belt, and the three-dimensional point cloud data of the laser arc in the belt surface image is calculated; Calculating the spatial distance value between each target point and the adjacent points in the three-dimensional point cloud data in the direction perpendicular to the tangent of the laser arc, and screening the points whose spatial distance value exceeds the preset distance threshold to form an initial candidate point set; Starting from each point in the initial candidate point set, the points that are adjacent in space and belong to the initial candidate point set are expanded through the breadth-first search algorithm to obtain the target candidate point set, so as to form multiple disconnected connected domains, and calculate the length and depth of each connected domain; Filtering the connected domains that do not meet the threshold condition among all connected domains, retaining the connected domains that meet the threshold condition, taking all the connected domains that meet the threshold condition as corresponding potential cracks, and forming a potential crack set with the spatial position information of all the potential cracks, wherein the non-meeting threshold condition means that both the length and the depth do not reach the corresponding preset threshold; During the belt circulation process, if there is a real crack, a belt tear detection result including the spatial position information of the real crack and the corresponding alarm level is generated. The real crack is a potential crack recorded at the same spatial position in the potential crack set in multiple consecutive operation cycles.

2. The method according to claim 1, characterized in that Starting from each point in the initial candidate point set, the points that are spatially adjacent and belong to the initial candidate point set are expanded by a breadth-first search algorithm to obtain a target candidate point set, including: Traversing each target point in the initial candidate point set, marking the target point as a member of the current connected domain, and inserting it into a preset queue to be viewed, and calculating the dynamic neighborhood range of the current target point according to the belt running direction and belt running speed; Searching for adjacent points in the dynamic neighborhood that are spatially adjacent and belong to the initial candidate point set along a direction perpendicular to the tangent of the laser arc; Determine whether the neighboring points of the target point belong to the initial candidate point set and are not marked; If yes, mark the adjacent point as a member of the current connected domain, insert it to the end of the preset queue to be checked, take the adjacent point as a new target point, and repeat the steps of marking, inserting, calculating, searching, and judging until all the target points in the preset queue to be checked form a connected domain, clear the preset queue to be checked, and repeat the process of forming a new connected domain until each target point in the initial candidate point set is marked; All target points that are not in any connected domain in the initial candidate point set are deleted to obtain a target candidate point set.

3. The method according to claim 2, characterized in that The judgment condition of spatial proximity is that the three-dimensional Euclidean distance between the adjacent point and the target point is less than a preset connection distance threshold, and the preset connection distance threshold is dynamically set according to the mapping relationship between the spot diameter of the line laser and the belt running speed calibrated by the experiment; The adjacent points in the search space and belonging to the initial candidate point set include: Calculate the tangent direction vector of the laser arc at the target point; Generate a vector perpendicular to the tangent direction of the laser arc according to the tangent direction vector of the laser arc and the unit vector of the belt thickness direction; The preset connected distance threshold is extended in the positive and negative directions of the tangent direction vector of the laser arc to form a strip-shaped search area, and points other than the target point in the strip-shaped search area are determined as adjacent points.

4. The method according to claim 1, characterized in that: The step of calculating the spatial distance between each target point and adjacent points in the three-dimensional point cloud data in a direction perpendicular to the tangent of the laser arc, and screening points whose spatial distance values ​​exceed a preset distance threshold to form an initial candidate point set includes: The cross-sectional data of the multi-layer composite structure is obtained through the preset belt cross-sectional scanning device, and the correlation model between the belt material and the belt thickness is established; According to the association model, dynamically generate a hierarchical distance threshold for each hierarchical structure where the target point is located; Calculate the spatial distance difference between the target point and the adjacent points in the anisotropic neighborhood in the belt thickness direction; compare the spatial distance difference with the stratified distance threshold, screen out the abnormal points whose spatial distance difference exceeds the stratified distance threshold, and determine the maximum penetration depth and stratified position information of each abnormal point; Constructing a density field based on the spatial distribution characteristics of the abnormal points, and calculating the spatial concentration of each abnormal point in the density field within a preset volume element; The product of the spatial concentration and the maximum penetration depth is determined as a comprehensive anomaly score, and abnormal points whose comprehensive anomaly scores exceed a preset score threshold are screened to form an initial candidate point set.

5. The method according to claim 4, characterized in that The determining of the maximum penetration depth and layered position information of each abnormal point includes: According to the spatial position information of the abnormal point, combined with the geometric model of the layer boundary, the maximum deviation between each abnormal point and the layer boundary is calculated to generate the maximum penetration depth; In combination with the hierarchical index table, hierarchical position analysis is performed on the spatial position information of the abnormal point to obtain hierarchical position information.

6. The method according to claim 1, characterized in that The spatial location information of all potential cracks is used to form a potential crack set, including: Extracting crack feature points according to the spatial positions of all the potential cracks and forming an initial potential crack point set; Performing spatial clustering on the initial potential fracture point set, and merging adjacent points to form potential fracture segments; Calculating the geometric features of all the potential fracture segments, analyzing the adjacency and intersection relationships between the fracture segments, and constructing a fracture topology map; Based on the geometric continuity and the fracture topology map, global optimization is performed to eliminate low-confidence fracture segments and obtain a potential fracture set.

7. The method according to claim 1, characterized in that The calculating the length and depth of each connected domain includes: Parsing the spatial positions and boundary information of all the connected domains to generate an initial description information set for each of the connected domains; According to the initial description information set and the preset spatial geometric relationship model, performing boundary extraction for each of the connected domains to obtain coordinates of all boundary points of each connected domain; According to the coordinates of all the boundary points of each of the connected domains, the relative position relationship between all the boundary points of the connected domains is analyzed, two first endpoints corresponding to the longest path are determined, and the distance value between the two first endpoints is used as the length of the connected domain; According to the coordinates of all boundary points of each connected domain, the distance distribution between all boundary points of the connected domain and the preset connected domain center point is analyzed, two second endpoints perpendicular to the longest path are determined, and the distance value between the two second endpoints is used as the depth of the connected domain.

8. A belt tear detection device based on three-dimensional vision, characterized in that: include: An acquisition module is used to acquire the belt surface image through a line laser and an industrial camera deployed under the belt, and calculate the three-dimensional point cloud data of the laser arc in the belt surface image; A screening module, used to calculate the spatial distance value between each target point and the adjacent points in the three-dimensional point cloud data in the direction perpendicular to the tangent of the laser arc, and screen the points whose spatial distance value exceeds a preset distance threshold to form an initial candidate point set; A calculation module is used to start from each point in the initial candidate point set, expand points that are adjacent in space and belong to the initial candidate point set through a breadth-first search algorithm, obtain a target candidate point set, so as to form multiple disconnected connected domains, and calculate the length and depth of each connected domain; A filtering module is used to filter the connected domains that do not meet the threshold condition in all connected domains, retain the connected domains that meet the threshold condition, take all the connected domains that meet the threshold condition as corresponding potential cracks, and form a potential crack set with the spatial position information of all potential cracks, wherein the non-meeting threshold condition means that both the length and the depth do not reach the corresponding preset threshold; A generation module is used to generate a belt tear detection result including spatial position information of the real crack and a corresponding alarm level if a real crack exists during the belt circulation operation. The real crack is a potential crack recorded at the same spatial position in the potential crack set in multiple consecutive operation cycles.

9. A computing device, characterized in that It comprises a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a belt tear detection method based on three-dimensional vision as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that: A computer program is stored, and when the computer program is executed by a computer, a belt tear detection method based on three-dimensional vision as described in any one of claims 1 to 7 is implemented.

Images