Belt conveyor deviation early warning method and system based on machine vision
By using edge detection and jitter decoupling methods based on machine vision technology, the problem of low detection accuracy of belt conveyor deviation caused by vertical jitter interference was solved, achieving high-precision deviation warning and reducing maintenance costs.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-04
- Publication Date
- 2026-03-10
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
During the operation of belt conveyors, vertical vibration interference leads to low accuracy in deviation detection. Existing contact detection methods suffer from low detection accuracy and high maintenance costs.
By using machine vision technology, the system acquires image frame sequences of the conveyor using a camera, performs left and right edge detection, and combines double edge difference and jitter decoupling technology to eliminate vertical jitter interference, accurately locate the center position of the conveyor belt, and achieve early warning of belt deviation.
It improves the accuracy and reliability of belt conveyor misalignment detection, reduces noise interference, enables accurate early warning of misalignment, and reduces maintenance costs.
Smart Images
Figure CN121639641A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of belt conveyor technology, and in particular to a belt conveyor deviation early warning method and system based on machine vision. Background Technology
[0002] During long-term operation, belt conveyors are prone to belt misalignment due to factors such as idler wear, installation errors, uneven tension, and improper load distribution. Belt conveyor misalignment not only causes problems like belt edge wear, material spillage, and equipment damage, but can also lead to safety accidents, impacting production efficiency and equipment lifespan. Traditional belt conveyor misalignment detection mainly uses contact-based methods such as misalignment switches and limit switches, but these methods suffer from low detection accuracy and high maintenance costs. With the development of machine vision technology, some research teams are dedicated to researching vision-based misalignment detection, primarily using cameras to monitor changes in the conveyor belt edge position in real time to determine the misalignment status. However, during actual operation, factors such as load, idler gaps, and equipment vibration cause vertical vibrations in the conveyor. These vertical vibrations and horizontal misalignment couple in the image, causing changes in the position of the conveyor belt edge in the image, severely interfering with the accuracy of misalignment detection. Therefore, how to effectively eliminate the interference of conveyor vibrations and improve the accuracy and reliability of misalignment detection has become an urgent problem to be solved. Summary of the Invention
[0003] This invention provides a machine vision-based method and system for early warning of belt conveyor deviation, in order to solve the problem that vertical vibration of the conveyor interferes with the accuracy of deviation detection.
[0004] To achieve the above objectives, in one aspect, the present invention provides a machine vision-based method for early warning of belt conveyor misalignment, the method comprising: By continuously acquiring images of a preset monitoring area of the belt conveyor using a camera at a preset frame rate, a continuous image frame sequence is obtained. Left and right edge detection is performed on each frame in the continuous image frame sequence to obtain left edge position sequence data and right edge position sequence data.
[0005] The left edge jitter feature data and the right edge jitter feature data are obtained by double edge difference jitter separation based on the left edge position sequence and the right edge position sequence.
[0006] The corrected center position of the conveyor belt is obtained by decoupling the vibration characteristics of the left and right edges and correcting the center position of the conveyor belt.
[0007] The conveyor belt center position after correction is compared with the preset conveyor center line to obtain the conveyor deviation amount, and deviation warning measures are implemented based on the conveyor deviation amount.
[0008] Furthermore, the method for obtaining left edge position sequence data and right edge position sequence data by performing left and right edge detection on each frame in a continuous image frame sequence includes: Each frame of the image is converted to grayscale to obtain the first image of the conveyor. The grayscale image is then blurred using a Gaussian kernel to obtain the second image of the conveyor. Finally, the edge of the second image of the conveyor is extracted using the Canny operator to obtain the third image of the conveyor.
[0009] The weighted average method is used to convert color images to grayscale images, simplifying the image data structure and thus reducing the amount of computation.
[0010] By using a 5×5 Gaussian kernel with a standard deviation of 1.2 for convolution operations, noise interference caused by particles on the conveyor belt is removed, while retaining the main edge structure information. This effectively reduces high-frequency noise caused by particles on the conveyor belt, thereby improving the stability and accuracy of edge detection.
[0011] The Canny edge detection operator is then used to extract edge contours from the image. The low threshold is set to 45 and the high threshold to 135. The left and right edges of the conveyor belt show clear white lines, thus accurately locating the boundary between the conveyor belt and the background.
[0012] A first detection area is set on the left side of the third image of the conveyor according to a preset area width. The edge pixels of the first detection area are extracted and the coordinates of the edge pixels are calculated to obtain the left edge position sequence data.
[0013] A detection window is set on the left side of the image to avoid interference caused by full-image search, thereby improving detection efficiency and accuracy. The edge position is determined by calculating the coordinates of all edge points within the detection area.
[0014] A second detection area is set on the right side of the third image of the conveyor according to a preset area width. The edge pixels of the second detection area are extracted and the coordinates of the edge pixels are calculated to obtain the right edge position sequence data.
[0015] This step converts visual image frames into sequential data, providing accurate edge location information for subsequent deviation analysis.
[0016] Furthermore, the method for obtaining left edge jitter feature data and right edge jitter feature data through dual-edge differential jitter separation based on the left edge position sequence and the right edge position sequence includes: The left edge position sequence and the right edge position sequence are respectively subjected to moving average filtering to obtain the left edge reference position sequence and the right edge reference position sequence.
[0017] The original jitter component of the left edge is obtained by calculating the difference between the left edge position sequence and the left edge reference position sequence, and the original jitter component of the right edge is obtained by calculating the difference between the right edge position sequence and the right edge reference position sequence.
[0018] Based on the original jitter components of the left and right edges, the jitter coupling coefficient matrix is obtained by performing principal component analysis on the two-dimensional jitter vector sequence using the first deviation analysis model.
[0019] Based on the jitter coupling coefficient matrix, the two-dimensional jitter vector sequence is decomposed into vertical jitter components using the second deviation analysis model. The left edge jitter feature data and the right edge jitter feature data are then calculated based on the vertical jitter components and the jitter coupling coefficient matrix.
[0020] Furthermore, the method for obtaining the jitter coupling coefficient matrix by performing principal component analysis on the two-dimensional jitter vector sequence based on the original jitter components of the left and right edges using the first deviation analysis model includes: The original jitter components on the left and right edges are combined frame by frame to construct a two-dimensional jitter matrix. The two-dimensional jitter matrix is then centered by subtracting the mean of each column.
[0021] Perform covariance analysis on the centered jitter matrix to obtain the covariance matrix, and perform eigenvalue decomposition on the covariance matrix to obtain the eigenvalues and the corresponding orthogonal eigenvector matrix.
[0022] The orthogonal eigenvector matrix is sorted according to the size of its eigenvalues and then used as the jitter coupling coefficient matrix.
[0023] Furthermore, the method for decomposing the two-dimensional jitter vector sequence into vertical jitter components using the second deviation analysis model based on the jitter coupling coefficient matrix, and calculating the left edge jitter feature data and right edge jitter feature data based on the vertical jitter components and the jitter coupling coefficient matrix includes: The eigenvector corresponding to the largest eigenvalue in the jitter coupling coefficient matrix is selected as the vertical jitter principal component direction vector.
[0024] The vertical jitter component for each frame is obtained by performing a vector dot product operation between each two-dimensional jitter vector in the two-dimensional jitter vector sequence and the direction vector of the vertical jitter principal component.
[0025] The left edge jitter feature data and the right edge jitter feature data are calculated based on the vertical jitter component and the jitter coupling coefficient matrix.
[0026] Furthermore, the method for obtaining the corrected conveyor belt center position by decoupling and correcting the conveyor belt center position based on the left edge jitter feature data and the right edge jitter feature data includes: Based on the left edge position sequence and the right edge position sequence, the first center position sequence is obtained as follows: .
[0027] in For time series indexing, It is the left edge position sequence. It is the right edge position sequence. It is the first central position sequence.
[0028] The second center position sequence is obtained by compensating the first center position sequence based on the left edge jitter feature data and the right edge jitter feature data.
[0029] The second center position sequence is fitted with a sliding window polynomial and the center position trend component is extracted as the corrected conveyor belt center position.
[0030] Furthermore, the method for compensating the first center position sequence based on the left edge jitter feature data and the right edge jitter feature data to obtain the second center position sequence includes: The left edge jitter feature data is paired frame-to-frame with the left edge position coordinates extracted from the original image, and the corrected left edge position sequence is obtained by subtracting from the original left edge pixel position frame by frame.
[0031] The right edge jitter feature data is paired frame-to-frame with the right edge position coordinates extracted from the original image, and the corrected right edge position sequence is obtained by subtracting from the original right edge pixel position frame by frame.
[0032] The corrected left edge position sequence and the corrected right edge position sequence are averaged frame by frame to obtain the center pixel position of each frame image, forming the second center position sequence.
[0033] Furthermore, the method of performing sliding window polynomial fitting on the second center position sequence and extracting the center position trend component as the corrected conveyor belt center position includes: A sliding window is set for the second center position sequence with a sliding interval of one image frame. Data points of the second center position are selected in each sliding window and fitted using the least squares method to obtain the fitting curve of the conveyor correction center.
[0034] The center position trend component of each frame is extracted from the center fitting curve of the conveyor correction as the corrected center position of the conveyor belt.
[0035] Furthermore, the method of comparing the corrected conveyor belt center position with the preset conveyor center line to obtain the conveyor deviation amount, and executing deviation warning measures based on the conveyor deviation amount, includes: Obtain the preset horizontal coordinate of the conveyor centerline, compare the corrected center position of the conveyor belt with the preset horizontal coordinate of the conveyor centerline, and calculate the difference to obtain the conveyor deviation.
[0036] The absolute value of the deviation amount sequence within a continuous time period is compared point by point with the preset deviation threshold. When the deviation amount of the conveyor exceeds the preset deviation amount threshold and the duration of continuous deviation reaches the preset duration threshold, a control signal is issued to provide a deviation warning.
[0037] Based on the same inventive concept, this invention provides a machine vision-based belt conveyor deviation early warning system, the system comprising an image acquisition module, a jitter analysis module, a correction module, and an early warning module connected in sequence.
[0038] The image acquisition module is used to continuously acquire images of a preset monitoring area of the belt conveyor through a camera at a preset frame rate to obtain a continuous image frame sequence. The left and right edges of each frame in the continuous image frame sequence are detected to obtain left edge position sequence data and right edge position sequence data.
[0039] The jitter parsing module is used to obtain left edge jitter feature data and right edge jitter feature data by double edge difference jitter separation based on the left edge position sequence and the right edge position sequence.
[0040] The correction module is used to obtain the corrected center position of the conveyor belt by decoupling and correcting the center position of the conveyor belt based on the left edge jitter feature data and the right edge jitter feature data.
[0041] The early warning module is used to compare the corrected center position of the conveyor belt with the preset center line of the conveyor to obtain the conveyor deviation amount, and to execute deviation early warning measures based on the conveyor deviation amount.
[0042] Compared with the prior art, the beneficial effects of this invention are: by combining the collaborative analysis of the left and right edge position sequences with the vertical jitter principal component direction vector analysis, the vertical jitter component is separated; by decoupling the vertical jitter features, image vibration artifacts are further eliminated; and by combining sliding window polynomial fitting to filter instantaneous interference noise, the center position is accurately corrected, thereby achieving accurate early warning of deviation when the vertical jitter of the conveyor interferes with the deviation detection. Attached Figure Description
[0043] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0044] Figure 1 This is a flowchart of the machine vision-based belt conveyor deviation early warning method according to Embodiment 1 of the present invention; Figure 2 This is a schematic diagram of the module composition of the machine vision-based belt conveyor deviation early warning system according to Embodiment 2 of the present invention. Detailed Implementation
[0045] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0046] Example 1: As Figure 1 As shown, this embodiment provides a machine vision-based method for early warning of belt conveyor misalignment. The method includes: S1. The camera continuously acquires images of the preset monitoring area of the belt conveyor at a preset frame rate to obtain a continuous image frame sequence. The left and right edges of each frame in the continuous image frame sequence are detected to obtain the left edge position sequence data and the right edge position sequence data.
[0047] For example, by continuously capturing images of the conveyor belt running at 1 frame per second using a camera with an image resolution of 1280×960 pixels, a total of 150 consecutive images are obtained. Each image contains the complete conveyor belt and its edge information, forming a continuous image frame sequence.
[0048] S2. Based on the left edge position sequence and the right edge position sequence, the left edge jitter feature data and the right edge jitter feature data are obtained by double edge difference jitter separation.
[0049] S3. Based on the left edge jitter feature data and the right edge jitter feature data, the center position of the conveyor belt is obtained by jitter decoupling and correcting the center position of the conveyor belt.
[0050] S4. Compare the corrected center position of the conveyor belt with the preset center line of the conveyor to obtain the conveyor deviation amount, and implement deviation warning measures according to the conveyor deviation amount.
[0051] It should be noted that the method for obtaining left edge position sequence data and right edge position sequence data by performing left and right edge detection on each frame in a continuous image frame sequence includes: Each frame of the image is converted to grayscale to obtain the first image of the conveyor. The grayscale image is then blurred using a Gaussian kernel to obtain the second image of the conveyor. Finally, the edge of the second image of the conveyor is extracted using the Canny operator to obtain the third image of the conveyor.
[0052] The weighted average method is used to convert color images to grayscale images, simplifying the image data structure and thus reducing the amount of computation.
[0053] By using a 5×5 Gaussian kernel with a standard deviation of 1.2 for convolution operations, noise interference caused by particles on the conveyor belt is removed while preserving the main edge structure information. For example, the edge region where the original gray value changed abruptly from 58 to 32 is transformed into a smooth transition from 58 to 32, which effectively reduces the high-frequency noise caused by particles on the conveyor belt, thereby improving the stability and accuracy of edge detection.
[0054] The Canny edge detection operator is then used to extract edge contours from the image. The low threshold is set to 45 and the high threshold to 135. The left and right edges of the conveyor belt show clear white lines, thus accurately locating the boundary between the conveyor belt and the background.
[0055] A first detection area is set on the left side of the third image of the conveyor according to a preset area width. The edge pixels of the first detection area are extracted and the coordinates of the edge pixels are calculated to obtain the left edge position sequence data.
[0056] A detection window is set on the left side of the image to avoid interference caused by full-image search, thereby improving detection efficiency and accuracy. The edge position is determined by calculating the coordinates of all edge points in the detection area. Note that the coordinates of the edge pixels in this patent refer to the horizontal coordinates of the image, which facilitates subsequent collaborative analysis of the left and right edge position sequence data to extract jitter features.
[0057] For example, a first detection area is set on the left side of the third image of the conveyor, with an abscissa range of 180-280 pixels and an area width of 100 pixels. This area completely covers the possible location range of the left edge. In the first frame, the abscissa of the left edge pixel is 247, in the second frame it is 241, and in the third frame it is 243. Similarly, all frames are extracted in sequence to obtain the complete left edge position sequence data.
[0058] A second detection area is set on the right side of the third image of the conveyor according to a preset area width. The edge pixels of the second detection area are extracted and the coordinates of the edge pixels are calculated to obtain the right edge position sequence data.
[0059] For example, a second detection area is set on the right side, with an abscissa range of 1420-1520 pixels and an area width of 100 pixels. In the first frame, the abscissa of the right edge pixel is 1476, in the second frame it is 1472, and in the third frame it is 1468. Similarly, all frames are extracted in sequence to obtain the complete right edge position sequence data.
[0060] This step converts visual image frames into sequential data, providing accurate edge location information for subsequent deviation analysis.
[0061] It should be noted that the method for obtaining left edge jitter feature data and right edge jitter feature data through double-edge difference jitter separation based on the left edge position sequence and the right edge position sequence includes: The left edge position sequence and the right edge position sequence are respectively subjected to moving average filtering to obtain the left edge reference position sequence and the right edge reference position sequence.
[0062] For example, by setting a moving average filter with a window length of 3 to filter the left edge position sequence [247, 241, 243] and the right edge position sequence [1476, 1472, 1468], the first data value of the left edge reference position is (247+241+243) / 3=243.67, and the first data value of the right edge reference position is (1476+1472+1468) / 3=1472. Similarly, all data values can be obtained to form the left edge reference position sequence and the right edge reference position sequence.
[0063] The original jitter component of the left edge is obtained by calculating the difference between the left edge position sequence and the left edge reference position sequence, and the original jitter component of the right edge is obtained by calculating the difference between the right edge position sequence and the right edge reference position sequence.
[0064] For example, taking the first three frames as an example, the original jitter components of the left edge are calculated as follows: 247-243.67=3.33, 241-243.67=-2.67, 243-243.67=-0.67. The original jitter components of the right edge are calculated as follows: 1476-1472=4, 1472-1472=0, 1468-1472=-4.
[0065] Based on the original jitter components of the left and right edges, the jitter coupling coefficient matrix is obtained by performing principal component analysis on the two-dimensional jitter vector sequence using the first deviation analysis model.
[0066] Based on the jitter coupling coefficient matrix, the two-dimensional jitter vector sequence is decomposed into vertical jitter components using the second deviation analysis model. The left edge jitter feature data and the right edge jitter feature data are then calculated based on the vertical jitter components and the jitter coupling coefficient matrix.
[0067] It should be noted that the method for obtaining the jitter coupling coefficient matrix by performing principal component analysis on the two-dimensional jitter vector sequence based on the original jitter components of the left and right edges using the first deviation analysis model includes: The original jitter components on the left and right edges are combined frame by frame to construct a two-dimensional jitter matrix. The two-dimensional jitter matrix is then centered by subtracting the mean of each column.
[0068] For example: The original jitter components from the left and right edges are combined frame by frame to construct a 150×2 two-dimensional jitter matrix. Taking the first three frames as an example, the matrix data is as follows: Each row represents the original jitter components at the left and right edges of a frame. The mean value of the left jitter is (3.33-2.67-0.67) / 3=0, and the mean value of the right jitter is (4+0-4) / 3=0. Therefore, taking the first three frames as an example, the centered jitter matrix is the same as the two-dimensional jitter matrix. .
[0069] Perform covariance analysis on the centered jitter matrix to obtain the covariance matrix, and perform eigenvalue decomposition on the covariance matrix to obtain the eigenvalues and the corresponding orthogonal eigenvector matrix.
[0070] The orthogonal eigenvector matrix is sorted according to the size of its eigenvalues and then used as the jitter coupling coefficient matrix.
[0071] For example: performing covariance analysis on the centered jitter matrix yields the covariance matrix. Eigenvalue decomposition yields the first eigenvalue, 20.35, and the corresponding orthogonal eigenvector is... The second eigenvalue is 1.87, and the orthogonal eigenvector corresponding to the second eigenvalue is... Furthermore, the jitter coupling coefficient matrix is obtained. .
[0072] It should be noted that the method for decomposing the two-dimensional jitter vector sequence into vertical jitter components using the second deviation analysis model based on the jitter coupling coefficient matrix, and calculating the left edge jitter feature data and right edge jitter feature data based on the vertical jitter components and the jitter coupling coefficient matrix includes: The eigenvector corresponding to the largest eigenvalue in the jitter coupling coefficient matrix is selected as the vertical jitter principal component direction vector.
[0073] The vertical jitter component for each frame is obtained by performing a vector dot product operation between each two-dimensional jitter vector in the two-dimensional jitter vector sequence and the direction vector of the vertical jitter principal component.
[0074] For example: Select the eigenvector corresponding to the largest eigenvalue of 20.35 from the jitter coupling coefficient matrix. As the direction vector of the vertical jitter principal component, the vertical jitter component corresponding to each frame is obtained by performing a vector dot product operation between each two-dimensional jitter vector in the two-dimensional jitter vector sequence and the direction vector of the vertical jitter principal component. This is frame 1. Frame 2: Frame 3: .
[0075] The left edge jitter feature data and the right edge jitter feature data are calculated based on the vertical jitter component and the jitter coupling coefficient matrix.
[0076] For example, taking the first three frames as an example, frame 1: the left edge jitter feature value is... The right edge jitter characteristic value is Similarly, the jitter feature value of the left edge of frame 2 can be calculated. The right edge jitter characteristic value is The jitter feature value of the left edge of frame 3 The right edge jitter characteristic value is Thus, the left edge jitter feature data is obtained as follows: The right edge jitter feature data is .
[0077] It should be noted that the method for obtaining the corrected conveyor belt center position by decoupling and correcting the conveyor belt center position based on the left edge jitter feature data and the right edge jitter feature data includes: Based on the left edge position sequence and the right edge position sequence, the first center position sequence is obtained as follows: .
[0078] in For time series indexing, It is the left edge position sequence. It is the right edge position sequence. It is the first central position sequence.
[0079] For example: the first center position of frame 1 is The first center position of frame 2 is The first center position of frame 3 is Therefore, the sequence of the first center positions in the first three frames is: .
[0080] The second center position sequence is obtained by compensating the first center position sequence based on the left edge jitter feature data and the right edge jitter feature data.
[0081] The second center position sequence is fitted with a sliding window polynomial and the center position trend component is extracted as the corrected conveyor belt center position.
[0082] It should be noted that the method for compensating the first center position sequence based on the left edge jitter feature data and the right edge jitter feature data to obtain the second center position sequence includes: The left edge jitter feature data is paired frame-to-frame with the left edge position coordinates extracted from the original image, and the corrected left edge position sequence is obtained by subtracting from the original left edge pixel position frame by frame.
[0083] For example: Frame-by-frame pairing of the left edge jitter feature data with the left edge position coordinates extracted from the original image, and frame-by-frame subtraction from the original left edge pixel positions to obtain the corrected left edge position values (Frame 1): Frame 2: Frame 3: Therefore, the corrected left edge position sequence is: .
[0084] The right edge jitter feature data is paired frame-to-frame with the right edge position coordinates extracted from the original image, and the corrected right edge position sequence is obtained by subtracting from the original right edge pixel position frame by frame.
[0085] For example: Frame-by-frame pairing of the right edge jitter feature data with the right edge position coordinates extracted from the original image, and frame-by-frame subtraction from the original right edge pixel positions to obtain the corrected right edge position values (Frame 1): Frame 2: Frame 3: Therefore, the corrected left edge position sequence is: .
[0086] The corrected left edge position sequence and the corrected right edge position sequence are averaged frame by frame to obtain the center pixel position of each frame image, forming the second center position sequence.
[0087] For example: Averaging the corrected left edge position sequence and the corrected right edge position sequence frame by frame yields the center pixel position of each frame. The second center position value for frame 1 is... The value of the second center position in frame 2 is The value of the second center position in frame 3 is Therefore, the sequence of the second center position in the first three frames is Similarly, the sequence of the second center positions of all frames can be obtained.
[0088] It should be noted that the method of performing sliding window polynomial fitting on the second center position sequence and extracting the center position trend component as the corrected conveyor belt center position includes: A sliding window is set for the second center position sequence with a sliding interval of one image frame. Data points of the second center position are selected in each sliding window and fitted using the least squares method to obtain the fitting curve of the conveyor correction center.
[0089] The center position trend component of each frame is extracted from the center fitting curve of the conveyor correction as the corrected center position of the conveyor belt.
[0090] For example: Set the sliding window to 3 pairs of second center position sequences. The fitting curve of the conveyor correction center was obtained by fitting using the least squares method. Further extract the trend component at the center position as follows: Frame 1: Frame 2: Frame 3: Then the trend component at the center position These are the corrected center positions of the conveyor belt for each frame.
[0091] It should be noted that the method of comparing the corrected conveyor belt center position with the preset conveyor center line to obtain the conveyor deviation amount, and executing deviation warning measures based on the conveyor deviation amount, includes: Obtain the preset horizontal coordinate of the conveyor centerline, compare the corrected center position of the conveyor belt with the preset horizontal coordinate of the conveyor centerline, and calculate the difference to obtain the conveyor deviation.
[0092] For example: when the preset horizontal coordinate of the conveyor centerline is 860, the conveyor deviation in the first frame is... The conveyor deviation in the second frame is The conveyor deviation in the third frame is The deviation sequence is .
[0093] The absolute value of the deviation amount sequence within a continuous time period is compared point by point with the preset deviation threshold. When the deviation amount of the conveyor exceeds the preset deviation amount threshold and the duration of continuous deviation reaches the preset duration threshold, a control signal is issued to provide a deviation warning.
[0094] For example: when the preset deviation threshold is When the preset duration threshold is 3 seconds, the deviation sequence is compared point by point with the preset deviation threshold. absolute value All greater than If the deviation persists for 3 seconds, a control signal will be issued to warn of deviation.
[0095] Example 2: Based on the same inventive concept, such as Figure 2 As shown, this embodiment provides a machine vision-based belt conveyor deviation early warning system, which includes an image acquisition module, a jitter analysis module, a correction module, and an early warning module connected in sequence.
[0096] The image acquisition module is used to continuously acquire images of a preset monitoring area of the belt conveyor through a camera at a preset frame rate to obtain a continuous image frame sequence. The left and right edges of each frame in the continuous image frame sequence are detected to obtain left edge position sequence data and right edge position sequence data.
[0097] The jitter parsing module is used to obtain left edge jitter feature data and right edge jitter feature data by double edge difference jitter separation based on the left edge position sequence and the right edge position sequence.
[0098] The correction module is used to obtain the corrected center position of the conveyor belt by decoupling and correcting the center position of the conveyor belt based on the left edge jitter feature data and the right edge jitter feature data.
[0099] The early warning module is used to compare the corrected center position of the conveyor belt with the preset center line of the conveyor to obtain the conveyor deviation amount, and to execute deviation early warning measures based on the conveyor deviation amount.
[0100] It should be noted that the specific methods by which each module performs operations in the system described in the above embodiments have been described in detail in Embodiment 1 of the method, and will not be elaborated here.
[0101] Finally, it should be noted that although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A machine vision-based belt conveyor deviation early warning method, characterized in that, The method comprises: obtaining a continuous image frame sequence by continuously acquiring images of a preset monitoring area of a belt conveyor through a camera at a preset frame rate; obtaining a left edge position sequence data and a right edge position sequence data by performing left and right edge detection on each frame of the continuous image frame sequence; obtaining left edge jitter feature data and right edge jitter feature data by double-edge differential jitter separation according to the left edge position sequence and the right edge position sequence; obtaining a corrected conveyor belt center position by jitter decoupling and correcting the conveyor belt center position according to the left edge jitter feature data and the right edge jitter feature data; 2. The machine vision-based belt conveyor misalignment early warning method according to claim 1, characterized in that, comparing the corrected conveyor belt center position with a preset conveyor center line to obtain a conveyor deviation amount, and performing a deviation warning measure according to the conveyor deviation amount. The method for obtaining a left edge position sequence data and a right edge position sequence data by performing left and right edge detection on each frame of the continuous image frame sequence comprises: performing grayscale processing on each frame of the image to obtain a first conveyor image, performing blur processing on the grayscale image through a Gaussian kernel to obtain a second conveyor image, and performing edge extraction on the second conveyor image through a Canny operator to obtain a third conveyor image; setting a first detection area on the left side of the third conveyor image according to a preset area width, extracting edge pixel points of the first detection area, and calculating coordinates of the edge pixel points to obtain the left edge position sequence data; 3. The machine vision-based belt conveyor misalignment early warning method according to claim 1, wherein, setting a second detection area on the right side of the third conveyor image according to the preset area width, extracting edge pixel points of the second detection area, and calculating coordinates of the edge pixel points to obtain the right edge position sequence data. The method for obtaining left edge jitter feature data and right edge jitter feature data by double-edge differential jitter separation according to the left edge position sequence and the right edge position sequence comprises: performing moving average filtering on the left edge position sequence and the right edge position sequence to obtain a left edge reference position sequence and a right edge reference position sequence; calculating a difference value between the left edge position sequence and the left edge reference position sequence to obtain a left edge original jitter component, and calculating a difference value between the right edge position sequence and the right edge reference position sequence to obtain a right edge original jitter component; performing principal component analysis on a two-dimensional jitter vector sequence through a first deviation analysis model according to the left edge original jitter component and the right edge original jitter component to obtain a jitter coupling coefficient matrix; 4. The machine vision-based belt conveyor misalignment early warning method according to claim 3, characterized in that, decomposing the two-dimensional jitter vector sequence through a second deviation analysis model according to the jitter coupling coefficient matrix to obtain a vertical jitter component, and calculating the left edge jitter feature data and the right edge jitter feature data according to the vertical jitter component and the jitter coupling coefficient matrix. The method for obtaining a jitter coupling coefficient matrix by performing principal component analysis on a two-dimensional jitter vector sequence through a first deviation analysis model according to a left edge original jitter component and a right edge original jitter component comprises: combining the left edge original jitter component and the right edge original jitter component to construct a two-dimensional jitter matrix, and performing centering processing on the two-dimensional jitter matrix by subtracting the mean value of each column from the column to obtain a centered jitter matrix; performing covariance analysis on the centered jitter matrix to obtain a covariance matrix, and performing eigenvalue decomposition on the covariance matrix to obtain eigenvalues and a corresponding orthogonal eigenvector matrix; The orthogonal eigenvector matrix is sorted according to the eigenvalue size and taken as a jitter coupling coefficient matrix.
5. The machine vision-based belt conveyor misalignment early warning method according to claim 3, wherein, The method for obtaining the left edge jitter characteristic data and the right edge jitter characteristic data according to the vertical jitter component and the jitter coupling coefficient matrix comprises the following steps: An eigenvector corresponding to a maximum eigenvalue is selected from the jitter coupling coefficient matrix as a vertical jitter principal component direction vector; Each two-dimensional jitter vector in the two-dimensional jitter vector sequence is subjected to a vector inner product operation with the vertical jitter principal component direction vector to obtain a corresponding vertical jitter component of each frame; The left edge jitter characteristic data and the right edge jitter characteristic data are calculated according to the vertical jitter component and the jitter coupling coefficient matrix.
6. The machine vision-based belt conveyor misalignment early warning method according to claim 5, wherein, The method for obtaining the modified conveyor belt center position according to the left edge jitter characteristic data and the right edge jitter characteristic data through jitter decoupling and correction of the conveyor belt center position comprises the following steps: A first center position sequence is obtained according to the left edge position sequence and the right edge position sequence as follows: ； wherein is a time series index, is a left edge position sequence, is a right edge position sequence, is a first center position sequence; The first center position sequence is compensated according to the left edge jitter characteristic data and the right edge jitter characteristic data to obtain a second center position sequence; The second center position sequence is subjected to sliding window polynomial fitting, and a center position trend component is extracted as the modified conveyor belt center position.
7. The machine vision-based belt conveyor misalignment early warning method of claim 6, wherein, The method for compensating the first center position sequence according to the left edge jitter characteristic data and the right edge jitter characteristic data to obtain the second center position sequence comprises the following steps: The left edge jitter characteristic data is frame-by-frame paired with the left edge position coordinates extracted from the original image, and a modified left edge position sequence is obtained by subtracting the original left edge pixel position by frame; The right edge jitter characteristic data is frame-by-frame paired with the right edge position coordinates extracted from the original image, and a modified right edge position sequence is obtained by subtracting the original right edge pixel position by frame; The modified left edge position sequence and the modified right edge position sequence are averaged by frame to obtain a center pixel position of each frame of image to form the second center position sequence.
8. The machine vision-based belt conveyor misalignment early warning method of claim 6, wherein, The method for performing sliding window polynomial fitting on the second center position sequence and extracting a center position trend component as the modified conveyor belt center position comprises the following steps: A sliding window is set for the second center position sequence at a sliding interval of one image frame, and a conveyor correction center fitting curve is obtained by fitting the second center position data points in each sliding window through a least square method; A center position trend component corresponding to each frame is extracted from the conveyor correction center fitting curve as the modified conveyor belt center position.
9. The machine vision-based belt conveyor misalignment early warning method of claim 1, wherein, The method for obtaining the conveyor runout amount by comparing the modified conveyor belt center position with a preset conveyor center line and executing a runout warning measure according to the conveyor runout amount comprises the following steps: A preset conveyor center line abscissa is obtained, the modified conveyor belt center position is compared with the preset conveyor center line abscissa, and a difference value is obtained to obtain the conveyor runout amount; and A runout warning measure is executed according to the conveyor runout amount. The absolute values of the deviation sequence in a continuous time period are compared with preset deviation thresholds point by point, and a control signal is sent out for deviation early warning when the conveyor deviation is greater than the preset deviation threshold and the continuous deviation duration reaches the preset duration threshold.
10. A machine vision based belt conveyor misalignment early warning system for performing the method of any one of claims 1-9, characterized in that, The system comprises an image acquisition module, a jitter analysis module, a correction module and an early warning module connected in sequence. The image acquisition module is configured to continuously acquire images of a preset monitoring area of the belt conveyor at a preset frame rate through a camera to obtain a continuous image frame sequence, and perform left and right edge detection on each image in the continuous image frame sequence to obtain left edge position sequence data and right edge position sequence data. The jitter analysis module is configured to obtain left edge jitter feature data and right edge jitter feature data through double-edge differential jitter separation according to the left edge position sequence and the right edge position sequence. The correction module is configured to obtain a corrected conveyor belt center position through jitter decoupling and correction of the conveyor belt center position according to the left edge jitter feature data and the right edge jitter feature data. The early warning module is configured to compare the corrected conveyor belt center position with a preset conveyor center line to obtain a conveyor deviation, and perform a deviation early warning measure according to the conveyor deviation.