Conveyor belt thickness measuring method, device and equipment based on 3D camera and storage medium

Abstract

The invention discloses a conveying belt thickness measuring method and device based on a 3D camera and a storage medium, and relates to the field of image processing. A 3D camera is used for collecting point cloud data, edges are searched inwards from the edge of a conveying belt roller, and an effective conveying belt area is determined according to the belt edges searched on the two sides; height data are extracted from roller areas which are not covered on the two sides, abnormal values are removed based on the quartile distance, and a roller datum plane is generated; the datum plane serves as a reference zero point for thickness measurement; and calculating the thickness of the conveyor belt based on the difference between the reference plane and the surface height of the conveyor belt, comparing according to a thickness threshold, and recording the coordinates and timestamps of the abnormal points. According to the scheme, the 3D camera is used for collecting the point cloud data, determining the effective conveying belt area, generating the roller datum plane to serve as the reference zero point, calculating the thickness difference, recording the abnormal points and uploading data to support analysis, and the method has the advantages that non-contact, online and high-precision conveying belt thickness measurement is achieved, and therefore the maintenance plan is optimized.

Description

Technical Field

[0001] This application relates to the field of image processing, and in particular to a method, apparatus, device and storage medium for measuring the thickness of a conveyor belt based on a 3D camera. Background Technology

[0002] During long-term operation, conveyor belts may experience uneven thickness, localized wear, or even breakage due to material abrasion, tensile deformation, and joint fatigue. Failure to obtain timely and accurate information on the thickness changes across the entire conveyor belt surface can easily lead to sudden equipment malfunctions, resulting in production stoppages, economic losses, and even safety accidents.

[0003] The main methods for conveyor belt thickness measurement in related technologies include contact measurement and two-dimensional vision-based detection. Contact measurement usually requires stopping the machine and sampling at limited locations, making it impossible to achieve full-surface, real-time thickness monitoring, and it is also prone to causing secondary damage to the conveyor belt surface. Although two-dimensional vision-based detection methods can achieve online measurement, the measurement accuracy is greatly affected by factors such as lighting, angle, and color due to the lack of depth information in two-dimensional images, and it is difficult to cover the entire conveyor belt surface, making it impossible to achieve comprehensive acquisition and analysis of three-dimensional thickness data.

[0004] With the development of 3D vision technology, 3D cameras can acquire high-precision depth information of object surfaces, making it possible to reconstruct the surface morphology and calculate the dimensions of conveyor belts. However, 3D camera-based conveyor belt inspection systems are mostly used for surface defect or local deformation detection, and there is a lack of efficient and stable methods for real-time, high-precision measurement of the thickness of the entire conveyor belt surface. Therefore, there is an urgent need for a thickness measurement method that can use 3D cameras to quickly acquire, accurately calculate, and monitor the thickness of the entire conveyor belt surface in real time, in order to meet the needs of industrial sites for equipment operation safety and intelligent maintenance. Summary of the Invention

[0005] This application provides a method, apparatus, equipment, and storage medium for conveyor belt thickness measurement based on a 3D camera, which has the advantages of realizing non-contact, online, and high-precision conveyor belt thickness measurement, wear trend analysis, and life prediction, thereby optimizing maintenance plans.

[0006] On one hand, this application provides a method for measuring the thickness of a conveyor belt based on a 3D camera, the method comprising: Use a 3D camera to collect point cloud data, find the edge from the edge of the conveyor belt roller inward, and determine the effective conveyor belt area based on the belt edges found on both sides; Height data is extracted from the uncovered roller areas on both sides, outliers are removed based on the interquartile range, and a roller reference plane is generated; the reference plane serves as the zero point for thickness measurement. The thickness of the conveyor belt is calculated based on the difference between the height of the reference plane and the surface of the conveyor belt, and compared with the thickness threshold. The coordinates and timestamps of abnormal points are recorded.

[0007] Specifically, before edge finding, the ROI area is set according to the roller installation position and width; after acquiring the point cloud data, the scan line point cloud is selected from the two edges of the ROI area inward. The height Z values ​​in the line scan point cloud are compared. When the difference in height Z values ​​does not exceed a set threshold, the edge is directly searched from the edge of the conveyor belt roller inward. When the difference in height Z values ​​exceeds the set threshold, the slope of the point cloud data is compensated to obtain a compensated point cloud. The edge is then searched inward based on the compensated point cloud.

[0008] Specifically, the step of performing slope compensation on the point cloud data to obtain a compensated point cloud includes: Select n consecutive frames of scan line point clouds and calculate the roller slope of the i-th frame scan line point cloud. , means as follows:

[0009] and represent the average Z values ​​of several point clouds at both ends of the point cloud in the i-th frame scan line; This represents the number of points in the point cloud of the i-th scan line; Based on the slope of n rollers The average roller slope was calculated. ; The point cloud data after n frames is compensated based on the average roller slope, as shown below:

[0010] Among them and These represent the height Z values ​​before and after compensation, respectively. Indicates the first scan line Index points.

[0011] Specifically, the process of removing outliers based on interquartile range and generating a roller reference plane includes: List the height Z values ​​of the two roller regions and sort them in ascending order; Calculate the quartile values ​​Q1 and Q3 in the list, where Q1 is the 25th percentile value among m values ​​and Q3 is the 75th percentile value among m values. The interquartile range (IQR) and outlier limits are calculated based on the difference between the Q1 and Q3 values, as follows: IQR = Q3 - Q1 Lower bound = Q1 - 1.5 × IQR, Upper bound = Q3 + 1.5 × IQR Extracting values ​​from the list Point cloud data within the range, and average height The reference plane for the roller is determined.

[0012] Specifically, the step of tracing the conveyor belt edges inward from the edge of the conveyor belt rollers and determining the effective conveyor belt area based on the belt edges found on both sides includes: For each frame of scan line point cloud, use the sliding window method to search from both sides inward; the length of the searchable range of the sliding window is half the standard width; When the difference between the maximum and minimum z values ​​of the point cloud within the selected window is greater than 1 / 2 times the standard thickness of the conveyor belt, the points on both sides of the window are determined as the boundary points of one side of the belt. The width of the two belt edges that are found is checked, and the area formed by the two valid edges that pass the check is determined as the valid conveyor belt area.

[0013] Specifically, the width verification process includes: Obtain the belt point cloud data after edge finding, and determine the edge finding width based on the bilateral y-coordinates; When the edge search width is greater than the standard width of the conveyor belt and the difference is greater than a set threshold, it is compared with the valid edge of the previous frame; when the edge search width is greater than the valid edge width of the previous frame, it is determined to be a false edge, and the sliding window continues to search for new boundary points inward. When the edge width is greater than the standard width of the conveyor belt and the difference is not greater than a set threshold, belt edge shrinkage is performed to adjust the boundary to the standard width of the conveyor belt. When the edge width is not greater than the standard width of the conveyor belt, the area formed by the two valid edges that have passed the verification is determined as the valid conveyor belt area.

[0014] Specifically, the process of shrinking the belt edge and adjusting the boundary to the standard width includes: Extract all height Z values ​​within a set threshold width inward from the two belt edges, and calculate the difference between the maximum and minimum z values; When at least one belt edge has a difference greater than the preset value, the belt edge with the excessive difference is unilaterally contracted to the standard width of the conveyor belt to determine two valid edges. When the difference calculated between the two belt edges is not greater than the preset value, the two belt edges are shrunk to the standard width of the conveyor belt to determine the two effective edges.

[0015] On the other hand, this application provides a conveyor belt thickness measurement device based on a 3D camera, the device comprising: The determination module is used to collect point cloud data using a 3D camera, search for edges from the edge of the conveyor belt rollers inward, and determine the effective conveyor belt area based on the belt edges found on both sides. The generation module is used to extract height data from the uncovered roller areas on both sides, remove outliers based on the interquartile range, and generate a roller reference plane; the reference plane serves as the reference zero point for thickness measurement. The calculation module is used to calculate the thickness of the conveyor belt based on the difference between the height of the reference plane and the surface of the conveyor belt, compare it with the thickness threshold, and record the coordinates and timestamps of abnormal points.

[0016] In another aspect, this application provides a computer device including a processor and a memory, wherein the memory stores at least one instruction, at least one program, code set or instruction set, wherein the at least one instruction, the at least one program, the code set or instruction set is loaded and executed by the processor to implement the conveyor belt thickness measurement method based on a 3D camera described above.

[0017] In another aspect, this application provides a computer-readable storage medium storing at least one instruction, at least one program, code set, or instruction set, wherein the at least one instruction, the at least one program, the code set, or the instruction set is loaded and executed by a processor to implement the conveyor belt thickness measurement method based on a 3D camera described above.

[0018] The beneficial effects of the technical solution provided in this application include at least the following: using a 3D camera to acquire point cloud data directly provides three-dimensional depth information of the conveyor belt surface. This solves the problems of lacking depth information in two-dimensional images and the measurement accuracy being easily affected by environmental factors such as lighting, angle, and color.

[0019] Furthermore, this method establishes a stable measurement reference zero point by tracing the effective conveyor belt area inward from the edge of the conveyor belt rollers and extracting height data from the uncovered roller areas on both sides to generate a roller reference plane. This method of establishing the reference plane, combined with interquartile range elimination of outliers, improves the accuracy and robustness of thickness measurement and reduces the impact of environmental noise on the measurement results.

[0020] Ultimately, this method not only enables real-time calculation of conveyor belt thickness and recording of anomalies, but also supports historical thickness curve plotting, wear trend analysis, and lifespan prediction by uploading data to a database, thereby optimizing maintenance plans. This transforms conveyor belt maintenance from reactive to predictive, improving the safety, reliability, and economic efficiency of equipment operation. Attached Figure Description

[0021] Figure 1 This is a flowchart of a conveyor belt thickness measurement method based on a 3D camera provided in an embodiment of this application; Figure 2 A three-dimensional view of a feasible equipment installation layout is shown; Figure 3 This is a schematic diagram of the installation of the rollers and conveyor belt; Figure 4 This is a schematic diagram of a camera detecting the height of a conveyor belt; Figure 5 This is a schematic diagram showing the edge-finding error caused by burrs on the right edge of the conveyor belt; Figure 6 This is a schematic diagram illustrating the principle of window edge finding. Figure 7 This is a detailed algorithm flowchart for determining the effective conveyor belt area; Figure 8 This is a structural block diagram of a conveyor belt thickness measuring device based on a 3D camera provided in an embodiment of this application; Figure 9 A structural block diagram of a computer device provided in an exemplary embodiment of this application is shown. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0023] In this article, "multiple" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. The character " / " generally indicates that the preceding and following related objects have an "or" relationship.

[0024] Traditional conveyor belt thickness measurement technologies cannot achieve real-time thickness monitoring across the entire belt surface. Specifically, contact-based measurements require machine shutdown and are limited to sampling in localized areas, resulting in discontinuous monitoring. Two-dimensional vision-based methods lack depth information, are affected by ambient lighting conditions and camera angle variations, leading to unstable measurement accuracy, and the entire conveyor belt surface is difficult to cover. Consequently, complete three-dimensional thickness data cannot be obtained, impacting accurate assessment of conveyor belt wear and timely maintenance decisions.

[0025] For example, in open-pit mine conveyor belt systems, the abrasive properties of the ore continuously act on the conveyor belt surface, resulting in uneven wear. Due to drastic changes in ambient light, 2D vision-based inspection equipment cannot reliably acquire depth information under strong light conditions, leading to significant fluctuations in thickness measurements. Furthermore, only a limited area of ​​the entire conveyor belt width is inspected, ignoring wear at the edges. This makes it impossible to fully grasp the actual thickness distribution of the conveyor belt, potentially leading to sudden breakage accidents.

[0026] If the above problems are not addressed, inaccurate and incomplete conveyor belt thickness monitoring will lead to distorted assessments of equipment operating status and the inability to identify potential fault points in a timely manner. Consequently, conveyor belts may break in undetected weak areas, causing equipment downtime and material leaks, disrupting production continuity, and threatening personnel safety. Furthermore, the lack of historical thickness data accumulation and analysis hinders the development of maintenance plans, makes predictive maintenance difficult to implement, and increases unplanned downtime and maintenance costs.

[0027] Therefore, this application provides a method for measuring the thickness of a conveyor belt based on a 3D camera. Figure 1 The corresponding flowchart includes the following: S1. Use a 3D camera to collect point cloud data, find the edge from the edge of the conveyor belt roller inward, and determine the effective conveyor belt area based on the belt edges found on both sides. Point cloud data is acquired using a 3D camera. In practical applications, the 3D camera can be mounted above a conveyor belt to acquire its 3D point cloud data by continuously scanning the conveyor belt surface. For example, the camera can be fixed on a bracket and acquire data at a preset frequency. Alternatively, the camera can be mounted on a movable platform and moved along the length of the conveyor belt to cover a wider area. This acquired point cloud data contains geometric information about the conveyor belt surface and its surrounding environment, forming the basis for all subsequent thickness calculations.

[0028] Figure 2 A three-dimensional view of a feasible equipment installation layout is shown, with the specific design as follows: Hardware components: 3D camera, industrial mounting bracket, synchronous triggering module (linked with conveyor belt encoder), industrial control computer, display and alarm module, and data storage device.

[0029] Installation layout: The 3D camera is mounted on the head pulley of the conveyor belt at a preset height and angle to ensure full coverage of the conveyor belt; auxiliary light sources are placed on both sides of the camera to ensure uniform illumination for point cloud acquisition.

[0030] Data Acquisition Connection: The camera is connected to the industrial control computer via Gigabit Ethernet. The trigger module and encoder synchronously output trigger signals to achieve synchronous data acquisition according to the position of the conveyor belt. The sampling frequency is preferably 3000 lines / second, and the Z-axis measurement accuracy can reach ±0.5 mm.

[0031] Next, the conveyor belt is traced inwards from the edge of the roller, and the effective conveyor belt area is determined based on the belt edges found on both sides. After acquiring the point cloud data, the boundaries of the conveyor belt need to be identified. For example, the edge position can be initially determined by analyzing the changes in height or density in the point cloud data. One approach is to set a fixed width in the point cloud data and scan inwards from both sides. When the height change within the scanned area exceeds a certain preset value, that position is identified as the belt edge. Another approach is to process the depth image obtained by projecting the point cloud data using image processing algorithms, such as edge detection operators, to identify the contour line of the conveyor belt. After determining the belt edges on both sides, the area defined by these two edges is determined as the effective conveyor belt area, and subsequent thickness measurements will only be performed within this area.

[0032] This application uses the configuration of ROI regions as an example for illustration. Figure 3 This is a schematic diagram of the roller and conveyor belt installation. The red line represents the complete 3D scan range, i.e., the ROI area constructed along the length of the roller. The yellow boxes represent the left and right roller sections, which serve as a baseline for focused area analysis (typically, the conveyor belt edge does not extend beyond the roller edge), and the effective conveyor belt area is extracted from them.

[0033] S2. Extract height data from the uncovered roller areas on both sides, remove outliers based on interquartile range and generate a roller reference plane; the reference plane serves as the zero point for thickness measurement. Next, height data is extracted from the uncovered roller areas on both sides. Outliers are removed based on the interquartile range (IQR), and a roller reference plane is generated. The interquartile range (IQR) is a robust statistical measure of data dispersion. It reflects the distribution range of the middle 50% of the data by calculating the difference between the third quartile (Q3) and the first quartile (Q1), with the formula: IQR = Q3 - Q1. Its core advantage lies in its independence from extreme values ​​(outliers), providing a more stable description of the central tendency and dispersion of the data while eliminating relatively discrete values.

[0034] After determining the effective conveyor belt area, a stable reference zero point needs to be established. For example, a certain number of point cloud data points can be selected from the surface areas of the rollers not covered by the conveyor belt on both sides. The height values ​​of these point cloud data points are extracted to form a height dataset. To eliminate the influence of noise or random factors in the data, outliers can be removed from these height data. One method is to sort all height values ​​and then directly calculate their average or median as the reference height. Another method is to set a fixed height range and directly exclude data points outside this range. In this way, a numerical value representing the height of the roller surface can be obtained, thereby generating a stable roller reference plane, which is used as the reference zero point for thickness measurement.

[0035] S3. Calculate the conveyor belt thickness based on the difference between the height of the reference plane and the surface of the conveyor belt, compare it with the thickness threshold, and record the coordinates and timestamps of abnormal points.

[0036] Once the roller reference plane is generated, it becomes the unified reference for all thickness calculations. This means that the height value of any point on the conveyor belt surface will be compared with this reference plane to determine the actual thickness at that point. The conveyor belt thickness is calculated based on the difference between the height of this reference plane and the height of the conveyor belt surface. After obtaining the three-dimensional height information of each point on the conveyor belt surface, the Z-coordinate value of each point is subtracted from the height value of the roller reference plane mentioned above.

[0037] Figure 4 This diagram illustrates a camera detecting the height of a conveyor belt. The orange area represents the roller surface, which is the ideal reference zero point. Z0 in the diagram represents the distance from the 3D camera to the roller, while Zi represents the distance from the 3D camera to the belt. The difference between the two can be used to calculate the conveyor belt thickness. This setting ensures the relativity and consistency of thickness measurement, avoiding the impact of absolute height drift caused by changes in camera installation height or environmental conditions on thickness calculations.

[0038] Furthermore, the coordinates and timestamps of abnormal points are recorded based on a comparison with thickness thresholds. The calculated conveyor belt thickness data is compared in real time with preset thickness thresholds. For example, a minimum thickness threshold and a maximum thickness threshold can be set. When the thickness of a point is lower than the minimum threshold or higher than the maximum threshold, that point is identified as an abnormal point. At this time, the three-dimensional coordinates (X, Y, Z) of the abnormal point and the time information of the abnormality are automatically recorded. This real-time comparison and recording mechanism enables operators to identify potential problem areas of the conveyor belt.

[0039] Finally, thickness data and anomaly records are uploaded to a database, supporting historical thickness curve plotting, wear trend analysis, and lifespan prediction to optimize maintenance plans. All calculated thickness data and anomaly records are stored in a database. By processing historical data in the database, thickness variation curves of the conveyor belt over different time periods can be generated, showing its wear status. Based on this historical data, statistical methods or machine learning algorithms can be used to predict the wear trend of the conveyor belt and further assess its remaining service life. These analytical results provide a basis for developing maintenance plans, thereby transforming traditional reactive maintenance into predictive maintenance, improving the reliability and economy of equipment operation.

[0040] Compared to detection methods based on 2D vision, this method utilizes a 3D camera to acquire point cloud data, directly providing 3D depth information of the conveyor belt surface. This solves the problems of 2D images lacking depth information and measurement accuracy being easily affected by environmental factors such as lighting, angle, and color. In the example, even when lighting conditions change, the 3D camera can still stably acquire height data, ensuring the reliability of thickness calculation.

[0041] Furthermore, this method establishes a stable measurement reference zero point by tracing the effective conveyor belt area inward from the edge of the conveyor rollers and extracting height data from the uncovered roller areas on both sides to generate a roller reference plane. This method of establishing the reference plane, combined with interquartile range filtering to eliminate outliers, improves the accuracy and robustness of thickness measurement and reduces the impact of environmental noise on the measurement results. In the example, even if there is a small amount of dust or debris on the roller surface, the interquartile range method can effectively filter out these outliers, ensuring the accuracy of the reference plane.

[0042] Ultimately, this method not only enables real-time calculation of conveyor belt thickness and recording of anomalies, but also supports historical thickness curve plotting, wear trend analysis, and lifespan prediction by uploading data to a database, thereby optimizing maintenance plans. This transforms conveyor belt maintenance from reactive to predictive, improving the safety, reliability, and economic efficiency of equipment operation.

[0043] In some embodiments, when the 3D camera is installed at inconsistent distances from the rollers on the left and right, the height Z value in the point cloud data differs too much, resulting in inaccurate edge finding and inability to correctly determine the effective conveyor belt area.

[0044] Therefore, after acquiring the point cloud data, this application selects extremely short scan line point clouds from both edges of the ROI region inwards, ensuring that the selected scan line point clouds contain only roller point clouds. Then, the height Z values ​​in the scan line point clouds are compared.

[0045] When the height Z-value difference does not exceed the set threshold, it indicates that the installation position and detection accuracy meet the requirements, and the height Z-value difference of the point clouds on the left and right sides of the rollers is within an acceptable range. Therefore, the edge-finding operation can be performed directly without complex corrections. At this time, the system will directly call the preset edge-finding algorithm to search from the edge of the rollers into the interior of the conveyor belt to determine the boundary of the conveyor belt.

[0046] When the height Z value difference exceeds the set threshold and the difference exceeds the preset tolerance range, it indicates that there is an inconsistency in the distance between the left and right sides of the roller when the 3D camera is installed. At this time, it is necessary to perform slope compensation on the point cloud data and then find the edge inward after obtaining the compensated point cloud.

[0047] Slope compensation is a data correction technique used to eliminate systematic deviations in the height direction of point cloud data caused by camera mounting tilt, making the point cloud data more accurately reflect the true geometry of the object. This can be achieved by calculating the overall tilt angle or slope of the roller point cloud and then linearly or non-linearly adjusting the Z-coordinates of all point clouds based on this slope, so that the roller surface tends to be horizontal after compensation. Alternatively, it can be achieved by fitting a plane to the roller point cloud and then projecting all point clouds onto this fitted plane, or by correcting the Z-values ​​according to the equation of the fitted plane.

[0048] This application provides a method for slope compensation of point cloud data, including: A. Select n consecutive frames of scan line point clouds and calculate the roller slope of the i-th frame of scan line point cloud. , means as follows:

[0049] and represent the average Z values ​​of several point clouds at both ends of the point cloud in the i-th frame scan line; This represents the number of points in the point cloud of the i-th scan line; B. Based on the slope of n rollers The average roller slope was calculated. ; can be represented as:

[0050] C. Compensation is performed on the point cloud data after n frames based on the average roller slope, as shown below:

[0051] Among them and These represent the height Z values ​​before and after compensation, respectively. Indicates the first scan line Index points.

[0052] In a specific embodiment, the final slope compensation value can be obtained by averaging 100 consecutive lines (100 frames), where n is typically 100. For each scan line (point cloud of the rollers at both ends), the average Z-value of the last 10 points and the average Z-value of the first 10 points can be selected.

[0053] As a specific implementation method, when the system detects that the 3D camera is installed at inconsistent distances from the rollers on the left and right sides, causing tilt in the point cloud data, a slope compensation process can be initiated. The system continuously acquires 100 frames (n=100) of scan line point cloud data. For each of these 100 frames, the system extracts the average Z-value of the leftmost 10 points in the scan line point cloud of that frame as... And the average Z-value of the 10 point clouds on the far right as... Simultaneously, the total number N of scan line point clouds in that frame is acquired. Then, the roller slope of the i-th frame is calculated according to the formula. After calculating the roller slopes of these 100 frames, an arithmetic average is performed to obtain the average roller slope. Subsequently, for all newly acquired scan line point cloud data, the system will use this average roller slope for real-time compensation. Specifically, for each point in the scan line, its original height... The height will be corrected using a formula based on its index P in the scan line, thus obtaining the compensated height. In this way, subsequent edge finding and thickness calculations will be performed on accurate point cloud data that has undergone slope compensation.

[0054] The above technical solutions can significantly improve the accuracy of subsequent conveyor belt edge finding, effective conveyor belt area determination, and roller reference plane generation, thereby ensuring the accuracy and reliability of conveyor belt thickness measurement, avoiding measurement deviations caused by data skew, and providing a more solid data foundation for conveyor belt wear trend analysis and life prediction.

[0055] This application also provides a method for removing outliers using interquartile range and generating a roller reference plane, specifically including the following: 1. Compile a list of the height Z values ​​of the two roller areas and sort them in ascending order; 2. Calculate the quartile values ​​Q1 and Q3 in the list. Q1 is the 25th percentile value among m values, and Q3 is the 75th percentile value among m values; expressed as: Q1 = (m-1) × 0.25, Q3 = (m-1) × 0.75; 3. Calculate the interquartile range (IQR) and outlier limits based on the difference between Q1 and Q3 values, as shown below: IQR = Q3 - Q1; Lower bound = Q1 - 1.5 × IQR, Upper bound = Q3 + 1.5 × IQR; If the position is not an integer, linear interpolation is used: Q = data[int(pos)] + (pos - int(pos)) × (data[int(pos)+1] - data[int(pos)]). pos is the sort value.

[0056] 4. Extract values ​​from the list Point cloud data within the range, and average height It is determined to be the reference plane for the roller.

[0057] 5. Based on average height The thickness is calculated based on the difference in height between the thickness and the surface height of the conveyor belt.

[0058] in, This represents the thickness of point i on the conveyor belt. This represents the distance from the camera to point i on the conveyor belt surface (or after slope compensation).

[0059] The aforementioned filtering mechanism effectively eliminates abnormal height values ​​caused by factors such as roller surface defects, stains, or measurement noise, ensuring the purity and reliability of the data used for reference plane calculation. Finally, the height Z values ​​of these rigorously screened "normal" point cloud data are averaged, and the resulting average height value is determined as the roller reference plane. This reference plane serves as the zero point for measuring the thickness of the conveyor belt, and its accuracy directly affects the final thickness measurement result.

[0060] Through the above process, this solution can effectively and robustly generate a roller reference plane after extracting height data from the uncovered roller areas on both sides in the conveyor belt thickness measurement method. This avoids the problem of reference plane offset caused by local noise or defects, thus making the conveyor belt thickness measurement results more accurate and reliable, and providing a high-quality data foundation for subsequent wear trend analysis and life prediction.

[0061] In some embodiments described above in this application, an edge-finding method is proposed to determine the effective conveyor belt area by tracing the edge inward from the edge of the conveyor belt roller. However, in this process, factors such as burrs, stains, local deformation, or light interference at the conveyor belt edge can lead to inaccurate edge-finding results, thereby affecting the reliability and accuracy of the effective area. For example... Figure 5 A schematic diagram illustrating edge-finding errors caused by burrs on the right edge of the conveyor belt. To address this, this application provides a more refined method for determining the effective conveyor belt area through edge-finding, comprising the following steps: A. Use the sliding window method to search from both sides inwards for the point cloud of each scan line; Figure 6 This is a schematic diagram illustrating the principle of edge finding using a sliding window. For each frame of data [(y1,z1),(y2,z2), ...,(yn,zn)], the sliding window method (window length defaults to 4, configurable) is used. Searching from both sides inwards means starting from the two ends of the scan line point cloud (i.e., the left and right edges of the conveyor belt) and gradually moving the window towards the center of the conveyor belt. The searchable range of the sliding window is half the standard width (determined by the y-coordinate). This means the search range is limited to the area from the centerline of the conveyor belt to its edge, avoiding invalid searches outside the conveyor belt and thus improving edge finding efficiency and accuracy.

[0062] B. When the difference between the maximum and minimum z values ​​of the point cloud within the selected window is greater than 1 / 2 times the standard thickness of the conveyor belt, the points on both sides of the window are determined as the boundary points of one side of the belt. This criterion utilizes the height difference between the conveyor belt edge and the rollers or background. Conveyor belts typically have a certain thickness, resulting in a significant height difference between their edges and the supporting structure. When the sliding window moves to the conveyor belt edge, the point cloud within the window will simultaneously contain points from both the conveyor belt surface and the rollers (or background), causing a significant increase in the difference in z-values ​​(height values). Comparing this difference to half the standard conveyor belt thickness effectively identifies the true boundary of the conveyor belt. Half the standard conveyor belt thickness is an empirical threshold and can be adjusted to 0.3 or 0.7 times the standard thickness depending on the actual conveyor belt material, wear condition, and measurement accuracy requirements. Once the condition is met, the two outermost points of the window are considered the boundary points of that side of the belt.

[0063] C. Verify the width of the two belt edges found, and determine the area formed by the two valid edges that pass the verification as the valid conveyor belt area.

[0064] After initially locating the boundary points of the two belt sides, this method further verifies the width of the two belt edges. Width verification refers to checking the reasonableness of the conveyor belt width determined by these two edge points after they have been initially located. Its purpose is to eliminate erroneous edge-finding results caused by factors such as noise, burrs, stains, or local deformation, ensuring that the identified boundaries are true and valid. Verification may include comparison with the standard width of the conveyor belt, comparison with historical data, or further analysis of the local features of the boundary point cloud. Finally, the area formed by the two valid edges that pass verification is determined as the valid conveyor belt area. After width verification, if the two belt edges are considered reliable, the area defined by these two edges is confirmed as the valid conveyor belt area. This area is the core area for subsequent conveyor belt thickness measurement, ensuring that the thickness calculation is performed on the conveyor belt itself, avoiding the inclusion of edge interference or non-conveyor belt areas in the measurement range, thus guaranteeing the accuracy of the measurement results.

[0065] Figure 7 This is a detailed algorithm flowchart for determining the effective conveyor belt area. The width verification process includes: a. Obtain the belt point cloud data after edge finding [(x1,y1,z1), (x2,y2,z2), (x3,y3,z3)…(xm,ym,zm)], and determine the edge finding width |ym-y1| based on the bilateral y-coordinates; b. When the edge search width is greater than the standard width of the conveyor belt and the difference is greater than the set threshold, it is compared with the valid edge of the previous frame; when the edge search width is greater than the valid edge width of the previous frame, it is determined to be a false edge, and the sliding window continues to search for new boundary points inward. For example, if the detected edge width is greater than the standard width of the conveyor belt, and the difference exceeds a set threshold of 10mm, it indicates that there may be an erroneous edge finding result caused by external interference. In this case, the system will further compare the edge width of the current frame with the valid boundary width confirmed in the previous frame. If the current width not only exceeds the standard but is also larger than the valid width of the previous frame, it is highly suspected that the currently found boundary is a "false edge," meaning it is not the real edge of the conveyor belt but a misjudgment caused by factors such as burrs, stains, local deformation, or light interference. In this situation, the system will not accept the current edge finding result but will instruct the edge finding module to continue searching into the inner area of ​​the conveyor belt to find a more accurate true boundary. If found, the edge width calculation continues; if not found, the edge finding is considered a failure, and the frame data is invalid.

[0066] c. When the edge width is greater than the standard width of the conveyor belt, and the difference is not greater than the set threshold, belt edge shrinkage is performed to adjust the boundary to the standard width of the conveyor belt. On the other hand, when the edge width is greater than the standard width of the conveyor belt, and the difference is no greater than a set threshold of 10mm, this may indicate slight expansion of the conveyor belt, measurement error, or slight natural fluctuations at the edge. In this case, the system will adopt a "shrinkage" strategy to shrink the belt edge and adjust the boundary to the standard width of the conveyor belt. Specifically, the system will adjust the currently found two boundaries inward so that the final determined width matches the standard width of the conveyor belt. This adjustment can be a proportional inward shift of the two boundaries, or a priority adjustment of a certain boundary based on the actual situation, to ensure that the width of the finally determined effective conveyor belt area meets the standard.

[0067] d. When the width of the edge search is not greater than the standard width of the conveyor belt, the area formed by the two valid edges that have passed the verification is determined as the valid conveyor belt area.

[0068] Finally, when the edge-finding width is not greater than the standard width of the conveyor belt, it means that the edge-finding result is reliable or has been effectively adjusted. In this case, the system will accept the currently found two boundaries and formally define the area enclosed by these two boundaries as the "effective conveyor belt area" for subsequent thickness measurement.

[0069] This hierarchical judgment and processing logic enables the system to flexibly handle various edge-finding situations. It can effectively eliminate false edges caused by strong interference and accurately correct slight deviations, thereby ensuring the accuracy of subsequent conveyor belt thickness measurements. This width verification process is closely integrated with the aforementioned edge-finding method from the edge of the conveyor belt roller inwards, and verification is performed immediately after the initial edge finding, forming a closed-loop feedback mechanism.

[0070] In some of the solutions described above in this application, a width verification process is proposed to handle abnormal edge width and ensure the accuracy of the effective conveyor belt area. However, in its implementation, when the edge width is greater than the standard width of the conveyor belt but the difference is not greater than the set threshold, directly shrinking the belt edge may ignore the height changes of the edge area (such as burrs, stains or local deformation), resulting in insufficient boundary adjustment, affecting the reliable determination of the effective conveyor belt area, and thus reducing the overall accuracy of thickness measurement.

[0071] In response, this application further proposes a method for adjusting the boundary to a standard width by shrinking the belt edge, including the following steps: A. Extract all height Z values ​​within a set threshold width inward from the two belt edges, and calculate the difference between the maximum and minimum z values; After determining the initial boundaries of the conveyor belt, the system extends a preset distance from each belt edge into the conveyor belt, forming a narrow, strip-shaped area. This distance can be configured according to the actual application scenario and the characteristics of the conveyor belt; for example, it can be set as a percentage of the standard width of the conveyor belt or a fixed value in millimeters. Within this strip-shaped area, the height Z-values ​​of all point cloud data acquired by the 3D camera will be extracted as the basis for subsequent analysis.

[0072] The difference between the maximum and minimum z-values ​​is calculated to quantify the smoothness or variability of the conveyor belt surface within the extracted area. For example, if burrs, stain accumulation, or localized deformation exist in the area, the difference between the maximum and minimum Z-values ​​will be relatively large. This difference serves as an important indicator for assessing edge anomalies.

[0073] B. When at least one belt edge difference is greater than the preset value, the belt edge with the excessive difference is unilaterally contracted to the standard width of the conveyor belt to determine two effective edges. This step provides an intelligent boundary adjustment strategy. When the height Z-value difference of a certain belt edge region exceeds a preset threshold, it indicates that there may be a local anomaly affecting measurement accuracy on that edge. In this case, the system will only retract and adjust the belt edge with the anomaly, moving it inwards until the width of the entire conveyor belt conforms to the standard width. This unilateral retraction method can specifically handle problematic edges, avoiding unnecessary adjustments to normal edges, thereby improving the accuracy of boundary determination.

[0074] C. When the difference calculated between the two belt edges is not greater than the preset value, the two belt edges are shrunk to the standard width of the conveyor belt to determine the two effective edges.

[0075] This step is another boundary adjustment strategy, applicable when both belt edges are relatively flat, but the overall width is still slightly larger than the standard width. In this case, since there are no obvious abnormalities on either side, the system will simultaneously shrink both belt edges inward by an equal or proportional amount to achieve the standard width of the conveyor belt. This bilateral shrinkage method maintains the stability of the conveyor belt centerline, ensuring that the adjusted effective conveyor belt area is geometrically more symmetrical and reasonable.

[0076] For example, if a preset Z-value difference threshold is set to a certain value, and the Z-value difference in the left area is greater than this threshold while the Z-value difference in the right area is not, the system will determine that there is an anomaly on the left edge. In this case, the system will only adjust the left belt edge inward until the overall width of the conveyor belt reaches the standard width, while the right belt edge remains unchanged. Conversely, if the Z-value differences in both the left and right areas are not greater than the preset value, the system will determine that there are no obvious anomalies on either side. In this case, both the left and right belt edges will be adjusted inward by an equal amount simultaneously to make the conveyor belt width reach the standard width. In this way, the system can flexibly select a single-sided or double-sided contraction strategy based on the actual shape characteristics of the edges, thereby accurately determining the effective conveyor belt area.

[0077] Through the above technical solution, this application can intelligently select a single-sided or double-sided shrinkage strategy based on the actual height fluctuation of the conveyor belt edge area, thereby overcoming the problem of inaccurate boundary adjustment that may occur when traditional methods deal with abnormal situations such as edge burrs, stains, or local deformation. This solution achieves accurate identification of abnormal edges and makes targeted adjustments through quantitative analysis of the Z-value difference in local edge height, avoiding excessive intervention in normal edges. This significantly improves the reliability and accuracy of determining the effective conveyor belt area, providing a more accurate measurement benchmark for subsequent conveyor belt thickness calculation, and thus improving the overall accuracy and stability of the entire thickness measurement method. In some of the solutions described above in this application, height data is extracted from the roller area to generate a reference plane for thickness measurement. However, during this process, the roller point cloud data may be abnormal due to noise, installation deviation, or environmental interference. Without verification, this can lead to an inaccurate reference plane, thus affecting the accuracy and reliability of thickness measurement. Verification after edge finding on both sides is used to distinguish between normal edge finding on both sides and situations where the belt has deviated to one side and the edge finding is on the belt itself. The verification process specifically includes: A. Sort the height Z values ​​of the roller point cloud data at both ends of the roller in ascending order to determine the 10th percentile value Q1' and the 90th percentile value Q9'; Q1' = m' × 0.1, Q9' = m' × 0.9; if the position is not an integer, then take the integer int(pos); B. Calculate the mean values ​​of the data in [0, Q1'] and [Q9', m'] respectively. and And calculate the difference between the two means; C. If If the point cloud data of the roller on that side is normal, continue with bilateral comparison; otherwise, if the point cloud data of the roller on that side is abnormal, determine that the edge finding of that frame of data is invalid. D. When the point cloud data of both ends of the roller is judged to be normal, the point cloud data of both sides of the roller are compared. The verification after edge finding on both sides is used to distinguish between normal edge finding on both sides and the case where the belt deviates to one side and the edge finding on the belt is on the belt. The calculation is as follows: when If the edge finding is successful, the edge finding for that frame of data is considered invalid; otherwise, the edge finding for that frame of data is considered invalid. and These represent the left and right reference edges found by the roller, respectively. and This represents the width verification threshold. L is used for internal consistency verification of point cloud data on a single roller, and D is used for consistency verification between the reference edges of both rollers. It is typically set to... .

[0078] Through the above technical solution, this application introduces a multi-level verification mechanism for roller point cloud data after determining the effective conveyor belt area. First, by sorting the height Z-values ​​of the single-sided roller point cloud data in ascending order and calculating the mean difference between the 10th and 90th percentiles, the internal consistency of the data on that side can be effectively assessed, and abnormal data caused by local noise, deformation, or installation deviations can be promptly identified and eliminated. Second, after both roller data have passed single-sided verification, the height difference between the reference edges of the left and right rollers is further compared, which can effectively identify and avoid inconsistencies in the reference edges caused by factors such as conveyor belt misalignment and camera installation tilt. This series of verification steps, progressively advancing, ensures the high quality and reliability of the original point cloud data used to generate the roller reference plane, thereby significantly improving the accuracy of the reference plane. Finally, based on a more accurate reference plane, the accuracy and stability of conveyor belt thickness measurement are greatly improved, effectively avoiding misjudgments caused by inaccurate references, providing more reliable data support for conveyor belt wear trend analysis and life prediction, thereby optimizing maintenance plans and reducing equipment failure risks. Example 1: System Setup and Hardware Configuration Conveyor belt parameters Bandwidth: 1400 mm; Belt speed: 3.0 m / s; Standard thickness: 25 mm; Operating environment: Mining and mineral processing workshop with high dust concentration; 3D camera model: Vision 3D camera; Resolution: 2048 dots per line; Line frequency: 3000 lines / second; Z-axis accuracy: ±0.5 mm; Installation height: 500 mm from the surface of the conveyor belt; Installation angle: Perpendicular to the normal of the conveyor belt; Installation method: The industrial carbon steel bracket + silicone shock-absorbing pad is fixed under the conveyor frame; Synchronous Trigger Module The pulse signal is collected by the encoder on the tail pulley of the conveyor belt and converted into a camera trigger signal; Encoder resolution: 1024 pulses / revolution, error ≤ 0.1%; Industrial control computers Model: Advantech IPC-610 Industrial PC; CPU: Intel i7-12700; Memory: 16 GB DDR4; Storage: 512 GB SSD + 1 TB HDD; Interfaces: 2 Gigabit Ethernet ports, 4 USB 3.0 ports; Pre-installed with Ubuntu + C++ / Python data processing modules; Example 2: Initialization and Calibration ROI setting: Set the Y-axis range to 0~1500 mm in the software UI to ensure that the shooting covers the belt and the rollers at both ends; Spatial calibration: using a standard ceramic calibration plate with an accuracy of ±0.01 mm; Ten frames of calibration point cloud data were collected, and a reference coordinate system for the conveyor belt running plane was established by least squares plane fitting. Slope compensation calculation (sample data) If a certain frame has 1000 scan lines, the average Z-value of the first 10 lines is 98.2 mm, and the average Z-value of the last 10 lines is 99.0 mm, then: k=(99.0 98.2) / 1000=0.0008 mm / point The final slope K is output by taking the average of 100 consecutive frames. Example 3: Operation and Data Processing The actual sampling frequency is 80 lines / second; A single frame point cloud contains 2048(Y)×1000(X) points; The preprocessing module uses median filtering (window = 3×3) and the point cloud registration accuracy is 0.2 mm. Sliding window edge finding (sample parameters) Window length: 4 points; Search range: 350~700 mm on each side; Judgment threshold: 0.6 × 25 mm = 15 mm; If the maximum difference in Z values ​​within the window is greater than or equal to the threshold, then the two ends of the window are recorded as boundary points. Benchmark value validity verification The data length for the left idler roller area is n=200. After sorting, the following values ​​are taken: Q1 position pos1 = 200 × 0.1 = 20 → mean_1 = 100.2 mm Q9 position pos9 = 200 × 0.9 = 180 → mean_9 = 100.5 mm The difference is 0.3 mm, which is less than the allowable deviation of 0.5 mm, so it is considered valid. IQR anomaly removal example Combined data for both side idlers n=400: Q1=100.1 mm, Q3=100.8 mm, IQR=0.7 mm; Lower bound = 100.1 - 1.5 × 0.7 = 99.05 mm; Upper bound = 100.8 + 1.5 × 0.7 = 101.85 mm; Only retain data within the range [99.05, 101.85], and the mean. ; Example 4: Thickness Calculation and Alarm Thickness calculation at a certain measurement point ; Height after compensation at point P ; thickness: ; Threshold judgment Preset thickness threshold = 3 mm; Thickness at this point: 2.4 mm < 3 mm → Alarm triggered Example 5: Data Storage and Analysis All thickness data and alarm information are uploaded to a MySQL database via an industrial network, supporting: displaying historical thickness curves, exporting wear distribution maps, and automatically generating monthly health reports; The results of 30 days of continuous on-site testing showed that: The maximum error in thickness measurement is ≤ ±0.5 mm; Anomaly alarm accuracy rate ≥98%; Maintenance response time has been reduced by more than 50%; Compared with the prior art, the present invention has the following beneficial effects: Full coverage: By deploying multi-view 3D cameras, synchronous thickness measurement of the entire conveyor belt is achieved, avoiding the limitations of traditional spot checks or single-point inspections.

[0079] Non-contact high-precision measurement: Utilizes 3D cameras to acquire high-precision depth data, avoiding equipment wear and downtime requirements caused by contact measurement.

[0080] Real-time online monitoring: Thickness data can be acquired and analyzed synchronously during the operation of the conveyor belt to achieve real-time early warning of thickness anomalies.

[0081] Data traceability: Historical thickness data can be used for wear analysis, life prediction and maintenance optimization, enabling refined management of conveying equipment.

[0082] Strong anti-interference capability: Through deep data processing and multi-frame fusion, it effectively suppresses the impact of illumination, color and surface contamination on measurement accuracy.

[0083] Figure 8 This is a structural block diagram of a conveyor belt thickness measuring device based on a 3D camera provided in an embodiment of this application. The device includes: The determination module 810 is used to collect point cloud data using a 3D camera, search for edges from the edge of the conveyor belt roller inward, and determine the effective conveyor belt area based on the belt edges found on both sides. The generation module 820 is used to extract height data from the uncovered roller areas on both sides, remove outliers based on the interquartile range, and generate a roller reference plane; the reference plane serves as the reference zero point for thickness measurement. The calculation module 830 is used to calculate the thickness of the conveyor belt based on the difference between the height of the reference plane and the surface of the conveyor belt, compare it with the thickness threshold, and record the coordinates and timestamps of abnormal points.

[0084] It should be noted that the conveyor belt thickness measurement device based on a 3D camera provided in this embodiment is only an example illustrating the division of the above-mentioned functional modules / units. In practical applications, the above functions can be assigned to different functional modules / units as needed, that is, the internal structure of the conveyor belt thickness measurement device based on a 3D camera can be divided into different functional modules / units to complete all or part of the functions described above. Furthermore, the implementation method of the conveyor belt thickness measurement method based on a 3D camera provided in the above method embodiment and the implementation method of the conveyor belt thickness measurement device based on a 3D camera provided in this embodiment belong to the same concept. For details of the specific implementation process of the conveyor belt thickness measurement device based on a 3D camera provided in this embodiment, please refer to the above method embodiment, which will not be repeated here.

[0085] Figure 9This application provides a structural block diagram of a computer device according to an exemplary embodiment. The computer device can be a desktop computer, laptop computer, handheld computer, or cloud server, etc. The computer device may include, but is not limited to, a processor and memory. The processor and memory can be connected via a bus or other means. The processor can be a central processing unit (CPU) or other general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic device, a graphics processing unit (GPU), an embedded neural network processor (NPU) or other dedicated deep learning coprocessor, discrete gate or transistor logic devices, discrete hardware components, or combinations of the above types of chips.

[0086] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs, non-transitory computer-executable programs, and modules, such as the program instructions / modules corresponding to the methods in the above embodiments of this application. The processor executes various functional applications and data processing by running the non-transitory software programs, instructions, and modules stored in the memory, thereby implementing the methods in the above embodiments. The memory may include a program storage area and a data storage area, wherein the program storage area may store the operating system and at least one application program required for a function; the data storage area may store data created by the processor, etc. Furthermore, the memory may include high-speed random access memory and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, the memory may optionally include memory remotely located relative to the processor, and these remote memories can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0087] Those skilled in the art will understand that the structure shown in this embodiment does not constitute a limitation on the computer device, and may include more or fewer components than shown, or combine certain components, or use different component arrangements.

[0088] This application also discloses a computer-readable storage medium. Specifically, the computer-readable storage medium is used to store a computer program, which, when executed by a processor, implements the methods described in the above-described method embodiments. Those skilled in the art will understand that implementing all or part of the processes in the methods described above can be accomplished by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments described above. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk (HDD), or solid-state drive (SSD), etc.; the storage medium can also include combinations of the above types of memory.

[0089] This specific embodiment is merely an explanation of the present invention and is not intended to limit the invention. After reading this specification, those skilled in the art can make modifications to this embodiment without contributing any inventive step, but such modifications are protected by patent law as long as they are within the scope of the claims of the present invention.

Claims

1. A method for measuring the thickness of a conveyor belt based on a 3D camera, characterized in that, The method includes: Use a 3D camera to collect point cloud data, find the edge from the edge of the conveyor belt roller inward, and determine the effective conveyor belt area based on the belt edges found on both sides; Height data is extracted from the uncovered roller areas on both sides, outliers are removed based on the interquartile range, and a roller reference plane is generated; the reference plane serves as the zero point for thickness measurement. The thickness of the conveyor belt is calculated based on the difference between the height of the reference plane and the surface of the conveyor belt, and compared with the thickness threshold. The coordinates and timestamps of abnormal points are recorded.

2. The method according to claim 1, characterized in that, Before edge finding, the ROI area is set according to the roller installation position and width; after obtaining the point cloud data, the scan line point cloud is selected from the two edges of the ROI area inward; The height Z values ​​in the line scan point cloud are compared. When the difference in height Z values ​​does not exceed a set threshold, the edge is directly searched from the edge of the conveyor belt roller inward. When the difference in height Z values ​​exceeds the set threshold, the slope of the point cloud data is compensated to obtain a compensated point cloud. The edge is then searched inward based on the compensated point cloud.

3. The method according to claim 1, characterized in that, The step of performing slope compensation on point cloud data to obtain compensated point cloud includes: Select n consecutive frames of scan line point clouds and calculate the roller slope of the i-th frame scan line point cloud. , means as follows: and represent the average Z values ​​of several point clouds at both ends of the point cloud in the i-th frame scan line; This represents the number of points in the point cloud of the i-th scan line; Based on the slope of n rollers The average roller slope was calculated. ; The point cloud data after n frames is compensated based on the average roller slope, as shown below: Among them and These represent the height Z values ​​before and after compensation, respectively. Indicates the first scan line Index points.

4. The method according to claim 1, characterized in that, The process of removing outliers based on interquartile range and generating a roller reference plane includes: List the height Z values ​​of the two roller regions and sort them in ascending order; Calculate the quartile values ​​Q1 and Q3 in the list, where Q1 is the 25th percentile value among m values ​​and Q3 is the 75th percentile value among m values. The interquartile range (IQR) and outlier limits are calculated based on the difference between the Q1 and Q3 values, as follows: IQR = Q3 - Q1 Lower bound = Q1 - 1.5 × IQR, Upper bound = Q3 + 1.5 × IQR Extracting values ​​from the list Point cloud data within the range, and average height The reference plane for the roller is determined.

5. The method according to claim 1, characterized in that, The process of tracing the conveyor belt edges inward from the edge of the conveyor belt rollers and determining the effective conveyor belt area based on the belt edges found on both sides includes: For each frame of scan line point cloud, use the sliding window method to search from both sides inward; the length of the searchable range of the sliding window is half the standard width; When the difference between the maximum and minimum z values ​​of the point cloud within the selected window is greater than 1 / 2 times the standard thickness of the conveyor belt, the points on both sides of the window are determined as the boundary points of one side of the belt. The width of the two belt edges that are found is checked, and the area formed by the two valid edges that pass the check is determined as the valid conveyor belt area.

6. The method according to claim 5, characterized in that, The width verification process includes: Obtain the belt point cloud data after edge finding, and determine the edge finding width based on the bilateral y-coordinates; When the edge search width is greater than the standard width of the conveyor belt and the difference is greater than a set threshold, it is compared with the valid edge of the previous frame; when the edge search width is greater than the valid edge width of the previous frame, it is determined to be a false edge, and the sliding window continues to search for new boundary points inward. When the edge width is greater than the standard width of the conveyor belt and the difference is not greater than a set threshold, belt edge shrinkage is performed to adjust the boundary to the standard width of the conveyor belt. When the edge width is not greater than the standard width of the conveyor belt, the area formed by the two valid edges that have passed the verification is determined as the valid conveyor belt area.

7. The method according to claim 6, characterized in that, The process of shrinking the belt edge and adjusting the boundary to the standard width includes: Extract all height Z values ​​within a set threshold width inward from the two belt edges, and calculate the difference between the maximum and minimum z values; When at least one belt edge has a difference greater than the preset value, the belt edge with the excessive difference is unilaterally contracted to the standard width of the conveyor belt to determine two valid edges. When the difference calculated between the two belt edges is not greater than the preset value, the two belt edges are shrunk to the standard width of the conveyor belt to determine the two effective edges.

8. A conveyor belt thickness measuring device based on a 3D camera, characterized in that, The device includes: The determination module is used to collect point cloud data using a 3D camera, search for edges from the edge of the conveyor belt rollers inward, and determine the effective conveyor belt area based on the belt edges found on both sides. The generation module is used to extract height data from the uncovered roller areas on both sides, remove outliers based on the interquartile range, and generate a roller reference plane; the reference plane serves as the reference zero point for thickness measurement. The calculation module is used to calculate the thickness of the conveyor belt based on the difference between the height of the reference plane and the surface of the conveyor belt, compare it with the thickness threshold, and record the coordinates and timestamps of abnormal points.

9. A computer device, characterized in that, The computer device includes a processor and a memory, the memory storing at least one instruction, at least one program, code set, or instruction set, the at least one instruction, the at least one program, the code set, or the instruction set being loaded and executed by the processor to implement the conveyor belt thickness measurement method based on a 3D camera as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The readable storage medium stores at least one instruction, at least one program, code set, or instruction set, wherein the at least one instruction, the at least one program, the code set, or the instruction set is loaded and executed by a processor to implement the 3D camera-based conveyor belt thickness measurement as described in any one of claims 1 to 7.

Images