Double-source image-based belt conveyor unbalance loading fault detection method

Abstract

The invention relates to the technical field of industrial equipment intelligent monitoring, in particular to a belt conveyor unbalance loading fault detection method based on a double-source image, which comprises the steps that an information acquisition module, a double-source image detection module and a central processing module are integrated on a belt conveyor, and the double-source image detection module comprises an RGB camera and an infrared camera; the method comprises the following steps: establishing a multiple regression model through a central processing module to calculate a temperature gradient amplitude, calculating a temperature weight mask through a temperature weight mask function, thereby obtaining double-source enhanced image data, extracting edge features for superposition and fusion, and setting an unbalance loading fault alarm level and an alarm trigger condition of the belt conveyor according to an edge pixel proportion; the problems that according to an existing belt conveyor unbalance loading detection method, shutdown detection is needed, the benefits are affected, the anti-interference capacity of single sensor detection is poor, and the image collection frame rate setting is not scientific are solved, through double-source image dynamic processing and fusion, high-quality edge detection images are output to accurately reflect the unbalance loading state, and the detection comprehensiveness is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent monitoring of industrial equipment, and in particular to a method for detecting eccentric load faults of a belt conveyor based on dual-source images. Background Art

[0002] Belt conveyor is a reliable and efficient material transport equipment. Due to its good economy, high degree of automation, safety, and good continuity, it is widely used and plays an important role in many fields such as mining, civil engineering, chemical industry, metallurgy, machinery manufacturing, electricity, and agriculture. As a common material transportation equipment, belt conveyor has one of the common faults during its operation. The main causes of overloading failure include uneven material loading and aging of belt conveyor parts. These factors will cause abnormal wear of the belt conveyor, which will reduce its conveying capacity and production efficiency, and increase safety hazards in the working process. The existence of safety hazards will further aggravate the impact of overloading failure. Therefore, overloading monitoring and fault warning of belt conveyors are important guarantees for the normal completion of production activities. Currently, the existing belt conveyor overloading detection mainly relies on manual inspection or single sensor inspection.

[0003] However, practice has shown that the existing belt conveyor overload detection method has the following problems: manual inspection often requires the belt conveyor to stop or run at a low speed to facilitate the staff to conduct inspections based on their experience, but this greatly reduces the production economic benefits and the inspection pass rate; single sensor inspection uses instruments for inspection compared to manual inspection, which improves the success rate and accuracy of the inspection to a certain extent, but only one instrument cannot fully reflect the overload status of the belt conveyor. Chinese patent document CN119117587A discloses a conveyor belt tilt detection device and its detection method. The conveyor belt tilt detection device is a cylindrical pressure block that presses the edge of the conveyor belt when an overload fault occurs, which drives the impact protrusion on the turntable below to hit the metal block and make a sound to alert the staff. However, under the normal working environment of the belt conveyor, the noise is often large, and it is difficult to distinguish whether it is an alarm sound caused by overload or noise in the working condition. Chinese patent document CN119734961A discloses a mining conveyor belt pre-tensioning device and method based on visual inspection. Its image acquisition device is a high-speed camera, which is easily affected by dust in the mining environment. Furthermore, the frame rate of the image acquisition device is dynamically set based solely on the conveyor belt's operating speed, without considering the impact of other factors on the frame rate adjustment. This shows that fault diagnosis methods using single sensor data are susceptible to interference from external factors, resulting in inaccurate fault diagnosis results and difficulty identifying belt conveyor overload faults. Furthermore, the image sensor acquisition frame rate is mostly fixed and not dynamically adjusted in conjunction with other belt conveyor parameters.

[0004] Therefore, it is necessary to invent a belt conveyor overload fault detection method based on dual-source images to solve the above problems. Summary of the Invention

[0005] In order to solve the problems of existing belt conveyor overload detection methods, such as the need to stop the machine for detection, which affects efficiency, the poor anti-interference ability of single sensor detection, and the unscientific image acquisition frame rate setting, the present invention provides a belt conveyor overload fault detection method based on dual-source images.

[0006] The present invention is achieved by adopting the following technical solutions:

[0007] A belt conveyor eccentric load fault detection method based on dual-source images includes the following steps:

[0008] S10: Integrate an information acquisition module, a dual-source image detection module, and a central processing module on the belt conveyor. The dual-source image detection module includes an RGB camera and an infrared camera. The central processing module sends a start-of-work instruction to the belt conveyor to control the information acquisition module to start working. The information acquisition module is used to collect speed and load signals of the belt conveyor and transmit them to the central processing module. The central processing module uses the collected speed and load signal data to dynamically adjust the camera sampling frame rate of the dual-source image detection module to obtain dual-source image data.

[0009] S20: Projecting the dual-source image data collected by the dual-source image detection module into a coordinate system through the central processing module, establishing a multivariate regression model of temperature, infrared grayscale value, and operating parameters of the belt conveyor to calculate the temperature gradient amplitude, thereby obtaining enhanced infrared image data; calculating a temperature weight mask through a temperature weight mask function, and using the temperature weight mask to perform block-by-block adaptive dynamic enhancement on the image data collected by the RGB camera to obtain enhanced RGB image data; the enhanced infrared image data and the enhanced RGB image data constitute the dual-source enhanced image data;

[0010] S30: extracting edge features of the dual-source enhanced image data using an edge extraction algorithm, superimposing and fusing the edge features of the dual-source enhanced image data using an edge feature fusion function to generate a dual-source edge feature fusion graph, and splicing the dual-source enhanced image data to form four-channel image data;

[0011] S40: Divide the dual-source edge feature fusion image by the center line, and statistically calculate the percentage of the number of edge pixels on both sides of the center line to the total number of edge pixels, and set the eccentric load fault alarm level and alarm triggering conditions of the belt conveyor based on this.

[0012] Furthermore, the information acquisition module includes a speed sensor and a pressure sensor, and the speed sensor is used to collect and calculate the speed signal of the belt conveyor. , pressure sensor is used to collect the load signal of belt conveyor .

[0013] Furthermore, the camera frame rate of the dual-source image detection module is dynamically adjusted by a dynamic sampling frame rate function, and the dynamic sampling frame rate function is:

[0014] ;

[0015] Where:

[0016] Indicates the basic sampling frame rate;

[0017] Indicates the maximum speed allowed by the belt conveyor system;

[0018] Indicates the maximum load allowed by the belt conveyor system;

[0019] 、 They represent the empirical weight coefficients of signal data respectively.

[0020] Furthermore, step S20 specifically includes:

[0021] S21: The RGB camera and the infrared camera are rigidly fixed to the front end of the belt conveyor through a bracket to collect dual-source image data. The points in the coordinate system of the infrared camera are projected to the coordinate system of the RGB camera through the central processing module. A multivariate regression model of the temperature value, the infrared grayscale value, and the working parameters of the belt conveyor is established. The multivariate regression model is:

[0022] ;

[0023] Where:

[0024] Indicates the location The target temperature value of the pixel;

[0025] Indicates the location Infrared grayscale value of the pixel;

[0026] represents the regression parameters estimated by least squares fitting;

[0027] S22: Combined with the target temperature value obtained by multivariate regression model, the temperature gradient amplitude is first calculated by operator convolution , that is, the enhanced infrared image data is obtained, and then the temperature weight mask is calculated through the temperature weight mask function. The temperature weight mask function is:

[0028] ;

[0029] Where:

[0030] represents the slope of the transition region;

[0031] represents the gradient threshold;

[0032] S23: Convert the image data collected by the RGB camera into HSV image data and process the brightness channel separately. Perform independent histogram equalization enhancement on the brightness channel in each block. Dynamically and adaptively adjust the contrast limit parameters of each block by the temperature weight mask. , to ensure that the enhancement intensity can match the temperature characteristics of each regional block, the dynamic adaptive adjustment formula of the contrast limit parameter is:

[0033] ;

[0034] Where:

[0035] Indicates the minimum value of contrast limit;

[0036] Indicates the maximum value of contrast limit;

[0037] S24: Merge the regional blocks of all brightness channels, convert the enhanced HSV image data back into image data of the RGB camera, and obtain enhanced RGB image data.

[0038] Furthermore, step S30 specifically includes:

[0039] S31: Extracting temperature edge features of enhanced infrared image data using edge extraction algorithm and visual edge features of enhanced RGB image data ;

[0040] S32: Temperature edge features are fused by edge feature fusion function and visual edge features Perform superposition fusion to generate a dual-source edge feature fusion map. The edge feature fusion function is:

[0041] ;

[0042] Where:

[0043] Indicates the location Fusion edge features of pixels;

[0044] Represents the infrared image temperature edge feature weight;

[0045] Represents the visual edge feature weight of the RGB image;

[0046] S33: The enhanced infrared image data is spliced ​​with the enhanced RGB image data to form four-channel image data.

[0047] Furthermore, step S40 specifically includes:

[0048] S41: Divide the dual-source edge feature fusion image into a left and a right part using the center line as a dividing line, count the number of edge pixels in the left and right parts respectively, and calculate the percentage of the number of edge pixels in the left and right parts to the total number of edge pixels respectively;

[0049] S42: According to the percentage of edge pixels in the left and right parts of the dual-source edge feature fusion image to the total number of edge pixels, set the belt conveyor's overload fault alarm level, alarm triggering conditions, and response measures:

[0050] When the eccentric load fault alarm level is level 3, that is, yellow alarm, it is within the allowable eccentric load range. The alarm trigger condition is the difference in the percentage of edge pixels in the left and right parts of the dual-source edge feature fusion image to the total edge pixels. If the rate is greater than 4% and less than or equal to 10%, the response measure is to record the abnormality and increase the monitoring frequency;

[0051] When the eccentric load fault alarm level is level 2, that is, orange alarm, it is a minor eccentric load fault. The alarm trigger condition is the difference in the percentage of edge pixels in the left and right parts of the dual-source edge feature fusion image to the total edge pixels. If the rate is greater than 10% and less than or equal to 20%, the response measure is to reduce speed and conduct manual inspection;

[0052] When the eccentric load fault alarm level is level 1, that is, red alarm, it is a serious eccentric load fault. The alarm trigger condition is the difference in the percentage of edge pixels in the left and right parts of the dual-source edge feature fusion image to the total edge pixels. If it is greater than 20%, the response measure is to immediately shut down the machine for maintenance.

[0053] Furthermore, the central processing module is a computer.

[0054] Furthermore, the pressure sensor is an electronic belt scale.

[0055] Furthermore, the speed sensor is an incremental rotary encoder.

[0056] Furthermore, the belt conveyor, RGB camera, infrared camera, speed sensor, and pressure sensor are all electrically connected to a computer.

[0057] The present invention provides a belt conveyor eccentric load fault detection method based on dual-source images, which has the following advantages over the prior art:

[0058] 1. The present invention combines the dual-source image acquisition technology of RGB camera and infrared camera, and fuses the dual-source image data to improve the comprehensiveness and comprehensiveness of belt conveyor overload fault detection.

[0059] 2. The present invention dynamically adjusts the sampling frame rate of the dual-source camera to obtain dual-source image data, thereby improving the image data quality when the belt conveyor working conditions require more, and optimizing performance and saving resources when the belt conveyor working conditions require less.

[0060] 3. The present invention proposes a multivariate regression model of temperature, infrared grayscale value, and operating parameters of the belt conveyor, taking into account the influence of belt conveyor speed and load, and solving the problem that the traditional linear regression model does not consider the changes of the belt conveyor under different operating conditions.

[0061] 4. The present invention proposes a temperature weight mask function to calculate the temperature weight mask, dynamically and adaptively enhance the independent histogram of each area block in the brightness channel, and ensure that the enhancement intensity can match the temperature characteristics of each area block.

[0062] 5. The present invention proposes an edge feature fusion function to generate a dual-source edge feature fusion graph, statistically calculates the percentage of edge pixels on both sides of the line in the dual-source edge feature fusion graph to the total number of edge pixels, and sets the belt conveyor overload fault alarm level and alarm triggering conditions accordingly. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 It is the overall flow chart of the present invention.

[0064] Figure 2 It is a structural schematic diagram of the belt conveyor in the present invention.

[0065] Figure 3 This is an edge detection image of the belt conveyor during normal operation in Comparative Example 1 of the present invention.

[0066] Figure 4 This is an edge detection image of the belt conveyor during normal operation in Comparative Example 2 of the present invention.

[0067] Figure 5 This is an edge detection image of the belt conveyor during normal operation in Example 1 of the present invention.

[0068] Figure 6 This is an edge detection image of the belt conveyor in comparative example 3 of the present invention when a third-level overload fault occurs.

[0069] Figure 7This is an edge detection image of the belt conveyor in comparative example 4 of the present invention when a third-level overload fault occurs.

[0070] Figure 8 This is an edge detection image of the belt conveyor when a third-level overload fault occurs in Example 2 of the present invention.

[0071] In the figure: 1. Belt conveyor; 2. RGB camera; 3. Infrared camera; 4. Speed ​​sensor; 5. Pressure sensor; 6. Bracket. DETAILED DESCRIPTION

[0072] The following is a clear and complete description of the relevant technical solutions. Obviously, the embodiments described are only some embodiments, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0073] The present invention will be further explained below with reference to the accompanying drawings, embodiments and comparative examples:

[0074] Example 1

[0075] A belt conveyor eccentric load fault detection method based on dual-source images is shown in the attached figure. Figure 1 As shown, the following steps are included:

[0076] S10: Integrate an information acquisition module, a dual-source image detection module and a central processing module on the belt conveyor 1. The dual-source image detection module includes an RGB camera 2 and an infrared camera 3. Send a start-of-work instruction to the belt conveyor 1 through the central processing module to control the information acquisition module to start working. The information acquisition module is used to collect the speed and load signals of the belt conveyor 1 and transmit them to the central processing module. The central processing module uses the collected speed and load signal data to dynamically adjust the camera sampling frame rate of the dual-source image detection module to obtain dual-source image data.

[0077] As attached Figure 2 As shown, the information acquisition module includes a speed sensor 4 and a pressure sensor 5. The speed sensor 4 is used to collect and calculate the speed signal of the belt conveyor 1. The pressure sensor 5 is used to collect the load signal of the belt conveyor 1. The pressure sensor 5 uses an electronic belt scale, and an independent rigid frame is built at the bottom of the belt conveyor 1 to install the electronic belt scale, replacing the traditional roller bracket to measure and collect the load signal of the belt conveyor 1. The speed sensor 4 adopts an incremental rotary encoder coaxially mounted with the driving roller of the belt conveyor 1. The incremental rotary encoder housing is fixed on the roller bearing seat and directly connected to the driving roller shaft end through a coupling to collect and calculate the speed signal of the conveyor belt of the belt conveyor 1. The central processing module is a computer, and the belt conveyor 1, RGB camera 2, infrared camera 3, speed sensor 4, and pressure sensor 5 are all electrically connected to the computer.

[0078] The camera frame rate of the dual-source image detection module is dynamically adjusted by a dynamic sampling frame rate function, which is:

[0079] ;

[0080] By dynamically adjusting the sampling frame rate of the dual-source camera to obtain dual-source image data, the image data quality is improved when the working condition requirements of the belt conveyor 1 increase, and the performance is optimized to save resources when the working condition requirements of the belt conveyor 1 decrease.

[0081] S20: The dual-source image data collected by the dual-source image detection module is projected into a coordinate system by the central processing module, and a multivariate regression model of temperature, infrared grayscale value, and operating parameters of the belt conveyor 1 is established to calculate the temperature gradient amplitude to obtain enhanced infrared image data; a temperature weight mask is calculated by a temperature weight mask function, and the temperature weight mask is used to perform block-by-block adaptive dynamic enhancement on the image data collected by the RGB camera 2 to obtain enhanced RGB image data; the enhanced infrared image data and the enhanced RGB image data constitute the dual-source enhanced image data.

[0082] Step S20 specifically includes:

[0083] S21: As attached Figure 2 As shown, in order to align the image data of RGB camera 2 and the image data of infrared camera 3 in time and space, RGB camera 2 and infrared camera 3 are rigidly fixed to the front end of belt conveyor 1 through T-shaped bracket 6 to collect dual-source image data. RGB camera 2 and infrared camera 3 are located on the same horizontal line. The central processing module projects the points in the coordinate system of infrared camera 3 to the coordinate system of RGB camera 2, and ensures that belt conveyor 1 is in the center of the image. In order to solve the problem that the traditional linear regression model does not consider the changes of belt conveyor 1 under different working conditions, the influence of real-time speed and load of belt conveyor 1 is considered, and a multiple regression model of temperature value, infrared gray value and working condition parameters of belt conveyor 1 is established. The multiple regression model is:

[0084] ;

[0085] Where:

[0086] Indicates the location The target temperature value of the pixel;

[0087] Indicates the location Infrared grayscale value of the pixel;

[0088] S22: Combined with the target temperature value obtained by multivariate regression model, the temperature gradient amplitude is first calculated by operator convolution And generate a temperature change intensity map, that is, obtain the enhanced infrared image data, and then calculate the temperature weight mask through the temperature weight mask function and generate a temperature weight mask intensity map for result visualization and later data fusion. The temperature weight mask function is:

[0089] .

[0090] S23: To ensure that the enhancement intensity of the dual-source image data matches the temperature characteristics of each block, so that the contrast of the target area is greatly enhanced to highlight the details, and the contrast of the background area is slightly enhanced to suppress noise, the image data collected by RGB camera 2 is converted into HSV image data and the brightness channel is processed separately. The 10×10 pixel block method with a 70% overlap rate of adjacent blocks is used to enhance the independent histogram equalization of the brightness channel blocks. The contrast limit parameter of each block is dynamically and adaptively adjusted by the temperature weight mask. , to ensure that the enhancement intensity can match the temperature characteristics of each regional block, the dynamic adaptive adjustment formula of the contrast limit parameter is:

[0091] .

[0092] S24: Merge the regional blocks of all brightness channels, convert the enhanced HSV image data back into image data of RGB camera 2, output a standard RGB image containing three channels of red (R), green (G), and blue (B), and obtain enhanced RGB image data.

[0093] S30: extracting edge features of the dual-source enhanced image data through an edge extraction algorithm, superimposing and fusing the edge features of the dual-source enhanced image data through an edge feature fusion function, generating a dual-source edge feature fusion graph, and splicing the dual-source enhanced image data into four-channel image data.

[0094] Step S30 specifically includes:

[0095] S31: extracting the temperature edge features of the enhanced infrared image data obtained in step S22 by an edge extraction algorithm and the visual edge features of the enhanced RGB image data obtained in step S24 .

[0096] S32: In order to improve the robustness and accuracy of the belt conveyor 1 eccentric load fault detection, the edge feature fusion function is used to fusion the temperature edge feature. and visual edge features The dual-source edge feature fusion image is generated by overlay fusion. The visual edge features of the RGB image data are enhanced to dominate in order to retain the rich texture structure. The temperature edge features of the infrared image data are enhanced to supplement the edge structure. The white and gray pixels in the dual-source edge feature fusion image are the fused edge data. The edge feature fusion function is:

[0097] .

[0098] S33: The enhanced infrared image data obtained in step S22 is spliced ​​with the enhanced RGB image data obtained in step S24 to form four-channel image data [R, G, B, T] including three channels of red (R), green (G), and blue (B) and a temperature gradient (T) channel, thereby further improving the robustness and accuracy of the overload fault detection of the belt conveyor 1.

[0099] S40: Divide the dual-source edge feature fusion image by the center line, and calculate the percentage of the edge pixels on both sides of the center line to the total edge pixels, and set the overload fault alarm level and alarm triggering conditions of the belt conveyor 1 accordingly.

[0100] Step S40 specifically includes:

[0101] S41: Divide the dual-source edge feature fusion image generated in step S32 into two parts, left and right, with the center line as the dividing line, and count the number of edge pixels in the left part respectively. and the number of edge pixels on the right side , calculate the percentage of edge pixels in the left part to the total number of edge pixels The percentage of edge pixels in the right part to the total edge pixels .

[0102] S42: The percentage of the number of edge pixels in the left portion calculated in step S41 to the total number of edge pixels The percentage of edge pixels in the right part to the total edge pixels , the difference between the two is definitely worth it , and accordingly set the overload fault alarm level and alarm triggering conditions of belt conveyor 1:

[0103] As shown in Table 1, when the percentage difference between the number of edge pixels in the left and right parts of the dual-source edge feature fusion image accounts for the total number of edge pixels The value range is When , the belt conveyor 1 is within the allowable eccentric load range, and the eccentric load state is named "Level 3 (yellow) eccentric load fault". At this time, it is necessary to record the eccentric load fault abnormality and increase the monitoring frequency to continuously observe the eccentric load trend; when the percentage difference between the number of edge pixels in the left and right parts of the dual-source edge feature fusion image and the total number of edge pixels is The value range is When , the belt conveyor 1 is in the range of slight eccentric load fault, and the eccentric load state is named "secondary (orange) eccentric load fault". At this time, a deceleration instruction needs to be issued to the belt conveyor 1, and the eccentric load fault situation needs to be checked manually. When the difference in the percentage of edge pixels in the left and right parts of the dual-source edge feature fusion image to the total number of edge pixels is The value range is When the load is too low, the belt conveyor 1 is in the serious overload fault range, and the overload state is named "Level 1 (red) overload fault". At this time, an immediate stop command must be issued to the belt conveyor 1, and the wear of the belt, frame, and roller structure of the belt conveyor 1 must be checked. The operation can only be restarted after the worn parts are replaced in time.

[0104]

[0105] It should be noted that, in this embodiment, the dual-source image detection module obtains dual-source image data of the belt conveyor 1 at an angle higher than the horizontal, that is, an upward viewing angle of 20 degrees from the horizontal plane, and processes the dual-source image data according to the above steps.

[0106] The results show that the total number of pixels of the image detected by this embodiment is 1849716 pixels, the total number of edge pixels is 293493 pixels, the edge pixel ratio is 15.87%, and the edge pixels in the left part account for the percentage of the total edge pixels. The percentage of edge pixels in the right part to the total number of edge pixels is 50.95%. It is 49.05%, and the difference between the two is absolutely worth it. , the overload fault alarm level of belt conveyor 1 has not been reached, and belt conveyor 1 is operating normally.

[0107] Comparative Example 1

[0108] The difference between this comparative example and Example 1 is that: the RGB image data of the belt conveyor 1 is obtained by using the RGB camera 2 alone, and the RGB image data is preprocessed, converted into a grayscale image and Gaussian blur processing is applied, the contrast is adjusted to perform edge detection, and an image display window is created to save the final edge detection result of the separate RGB image. The white pixels in the image are the edge data detected by the separate RGB image. The height and angle of the RGB camera 2 relative to the belt conveyor 1 in this comparative example are the same as the height and angle of the RGB camera 2 relative to the belt conveyor 1 in Example 1. The infrared camera 3 is not involved in the entire process.

[0109] The results show that the total number of pixels of the image detected by RGB camera 2 alone is 1849716 pixels, the total number of edge pixels is 43080 pixels, the edge pixel ratio is 2.33%, and the edge pixels in the left part account for the percentage of the total edge pixels. The percentage of edge pixels in the right part to the total number of edge pixels is 51.67%. It is 48.33%, and the difference between the two is absolutely worth it. , the overload fault alarm level of belt conveyor 1 has not been reached, and belt conveyor 1 is operating normally.

[0110] Comparative Example 2

[0111] The difference between this comparative example and Example 1 is that: the infrared image data of the belt conveyor 1 is obtained by using the infrared camera 3 alone, and the infrared image data is preprocessed, converted into a grayscale image and Gaussian blur processing is applied, the contrast is adjusted to perform edge detection, and an image display window is created to save the final edge detection result of the separate infrared image. The white pixels in the image are the edge data detected by the separate infrared image. The height and angle of the infrared camera 3 relative to the belt conveyor 1 in this comparative example are the same as the height and angle of the infrared camera 3 relative to the belt conveyor 1 in Example 1, and the RGB camera 2 is not involved in the entire process.

[0112] The results show that the total number of pixels of the image detected by infrared camera 3 alone is 1849716 pixels, the total number of edge pixels is 21126 pixels, the edge pixel ratio is 1.14%, and the edge pixels in the left part account for the percentage of the total edge pixels. The percentage of edge pixels in the right part to the total number of edge pixels is 50.38%. It is 49.62%, and the difference between the two is absolutely worth it. , the overload fault alarm level of belt conveyor 1 has not been reached, and belt conveyor 1 is operating normally.

[0113]

[0114] Reference Attachment Figure 3 ~Attached Figure 5 As can be seen from Table 2, the edge pixel ratio of the image detected by the belt conveyor overload fault detection method based on dual-source images proposed in the present invention is increased by 13.54% compared with the edge pixel ratio of the image detected by the RGB camera 2 alone, and is increased by 14.73% compared with the edge pixel ratio of the image detected by the infrared camera 3 alone. The overload fault detection results of the belt conveyor 1 are consistent, which proves the effectiveness of the present invention.

[0115] Example 2

[0116] The difference between this embodiment and embodiment 1 is that an overload fault of the belt conveyor 1 is artificially created while ensuring safety, and the dual-source image detection module in this embodiment obtains dual-source image data of the belt conveyor 1 at a semi-overhead angle, that is, an overhead perspective at an angle of 45 degrees to the horizontal plane, and the remaining detection steps are the same.

[0117] The results show that the total number of pixels of the image detected by this embodiment is 370304 pixels, the total number of edge pixels is 119439 pixels, the edge pixel ratio is 32.56%, and the edge pixels in the left part account for the percentage of the total edge pixels. The percentage of edge pixels in the right part to the total number of edge pixels is 52.15%. It is 47.85%, and the difference between the two is absolutely worth it. , reaching the third level of overload fault alarm level of belt conveyor 1, it is necessary to record the overload fault abnormality and increase the monitoring frequency to continuously observe the overload trend.

[0118] Comparative Example 3

[0119] The difference between this comparative example and Example 2 is that: the RGB image data of the belt conveyor 1 is obtained by using the RGB camera 2 alone, and the RGB image data is preprocessed, converted into a grayscale image and Gaussian blur processing is applied, the contrast is adjusted to perform edge detection, and an image display window is created to save the final edge detection result of the separate RGB image. The white pixels in the image are the edge data detected by the separate RGB image. The height and angle of the RGB camera 2 relative to the belt conveyor 1 in this comparative example are the same as the height and angle of the RGB camera 2 relative to the belt conveyor 1 in Example 2. The infrared camera 3 is not involved in the entire process.

[0120] The results show that the total number of pixels of the image detected by RGB camera 2 alone is 370304 pixels, the total number of edge pixels is 12210 pixels, the edge pixel ratio is 3.33%, and the edge pixels in the left part account for the percentage of the total edge pixels. The percentage of edge pixels in the right part to the total number of edge pixels is 53.71%. It is 46.29%, and the difference between the two is absolutely worth it. , reaching the third level of overload fault alarm level of belt conveyor 1, it is necessary to record the overload fault abnormality and increase the monitoring frequency to continuously observe the overload trend.

[0121] Comparative Example 4

[0122] The difference between this comparative example and Example 2 is that: the infrared image data of the belt conveyor 1 is obtained by using the infrared camera 3 alone, and the infrared image data is preprocessed, converted into a grayscale image and Gaussian blur processing is applied, the contrast is adjusted to perform edge detection, and an image display window is created to save the final edge detection result of the separate infrared image. The white pixels in the image are the edge data detected by the separate infrared image. The height and angle of the infrared camera 3 relative to the belt conveyor 1 in this comparative example are the same as the height and angle of the infrared camera 3 relative to the belt conveyor 1 in Example 2, and the RGB camera 2 is not involved in the entire process.

[0123] The results show that the total number of pixels of the image detected by infrared camera 3 alone is 370304 pixels, the total number of edge pixels is 13445 pixels, the edge pixel ratio is 3.63%, and the edge pixels in the left part account for the percentage of the total edge pixels. The percentage of edge pixels in the right part to the total number of edge pixels is 52.10%. It is 47.90%, and the difference between the two is absolutely worth it. , reaching the third level of overload fault alarm level of belt conveyor 1, it is necessary to record the overload fault abnormality and increase the monitoring frequency to continuously observe the overload trend.

[0124]

[0125] Reference Attachment Figure 6 ~Attached Figure 8 As can be seen from Table 3, the edge pixel ratio of the image detected by the belt conveyor overload fault detection method based on dual-source images proposed in the present invention is increased by 29.23% compared with the edge pixel ratio of the image detected by using the RGB camera 2 alone, and is increased by 28.93% compared with the edge pixel ratio of the image detected by using the infrared camera 3 alone. The overload fault detection results of the belt conveyor 1 are consistent, which proves the effectiveness of the present invention.

[0126] From the results of various embodiments and comparative examples, it can be seen that the edge pixel ratio of the image detected by the belt conveyor overload fault detection method based on dual-source images proposed in the present invention is significantly improved, and higher quality edge detection images are effectively output. At the same time, it can correctly reflect the overload status of the belt conveyor and realize the overload fault detection of the belt conveyor.

[0127] In the description of the present invention, it should be understood that the indicated orientations or positional relationships are based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention.

[0128] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A belt conveyor eccentric load fault detection method based on dual-source images, characterized by: The following steps are involved: S10: Integrate an information acquisition module, a dual-source image detection module, and a central processing module on the belt conveyor (1), wherein the dual-source image detection module includes an RGB camera (2) and an infrared camera (3); send a start-of-work instruction to the belt conveyor (1) through the central processing module to control the information acquisition module to start working, and the information acquisition module is used to collect speed and load signals of the belt conveyor (1) and transmit them to the central processing module; the central processing module dynamically adjusts the camera sampling frame rate of the dual-source image detection module using the collected speed and load signal data to obtain dual-source image data; S20: The dual-source image data collected by the dual-source image detection module are projected into a coordinate system by the central processing module, and a multivariate regression model of temperature, infrared grayscale value, and working condition parameters of the belt conveyor (1) is established to calculate the temperature gradient amplitude, thereby obtaining enhanced infrared image data; The temperature weight mask is calculated by a temperature weight mask function, and the image data collected by the RGB camera (2) is adaptively and dynamically enhanced in blocks using the temperature weight mask to obtain enhanced RGB image data; the enhanced infrared image data and the enhanced RGB image data constitute dual-source enhanced image data; S30: extracting edge features of the dual-source enhanced image data using an edge extraction algorithm, superimposing and fusing the edge features of the dual-source enhanced image data using an edge feature fusion function to generate a dual-source edge feature fusion graph, and splicing the dual-source enhanced image data to form four-channel image data; S40: Divide the dual-source edge feature fusion image by the center line, and calculate the percentage of the edge pixels on both sides of the center line to the total edge pixels, and set the overload fault alarm level and alarm triggering conditions of the belt conveyor (1) accordingly.

2. The method for detecting eccentric load faults of a belt conveyor based on dual-source images according to claim 1, characterized in that: The information acquisition module includes a speed sensor (4) and a pressure sensor (5). The speed sensor (4) is used to collect and calculate the speed signal of the belt conveyor (1). , the pressure sensor (5) is used to collect the load signal of the belt conveyor (1) .

3. The method for detecting eccentric load faults of a belt conveyor based on dual-source images according to claim 2, characterized in that: The camera frame rate of the dual-source image detection module is dynamically adjusted by a dynamic sampling frame rate function, which is: ； Where: Indicates the basic sampling frame rate; Indicates the maximum speed allowed by the system for the belt conveyor (1); Indicates the maximum load allowed by the belt conveyor (1) system; 、 They represent the empirical weight coefficients of signal data respectively.

4. The method for detecting eccentric load faults of a belt conveyor based on dual-source images according to claim 2, characterized in that: Step S20 specifically includes: S21: The RGB camera (2) and the infrared camera (3) are rigidly fixed to the front end of the belt conveyor (1) through the bracket (6) to collect dual-source image data. The points in the coordinate system of the infrared camera (3) are projected to the coordinate system of the RGB camera (2) through the central processing module, and a multivariate regression model of the temperature value, the infrared gray value, and the working condition parameters of the belt conveyor (1) is established. The multivariate regression model is: ； Where: Indicates the location The target temperature value of the pixel; Indicates the location Infrared grayscale value of the pixel; represents the regression parameters estimated by least squares fitting; S22: Combined with the target temperature value obtained by multivariate regression model, the temperature gradient amplitude is first calculated by operator convolution , that is, the enhanced infrared image data is obtained, and then the temperature weight mask is calculated through the temperature weight mask function. The temperature weight mask function is: ； Where: represents the slope of the transition region; represents the gradient threshold; S23: Convert the image data collected by the RGB camera (2) into HSV image data and process the brightness channel separately. Perform independent histogram equalization enhancement on the brightness channel in each block. Dynamically and adaptively adjust the contrast limit parameters of each block by the temperature weight mask. , to ensure that the enhancement intensity can match the temperature characteristics of each regional block, the dynamic adaptive adjustment formula of the contrast limit parameter is: ； Where: Indicates the minimum value of contrast limit; Indicates the maximum value of contrast limit; S24: Merge the regional blocks of all brightness channels, convert the enhanced HSV image data back into the image data of the RGB camera (2), and obtain enhanced RGB image data.

5. The method for detecting eccentric load faults of a belt conveyor based on dual-source images according to claim 1, characterized in that: Step S30 specifically includes: S31: Extracting temperature edge features of enhanced infrared image data using edge extraction algorithm and visual edge features of enhanced RGB image data ; S32: Temperature edge features are fused by edge feature fusion function and visual edge features Perform superposition fusion to generate a dual-source edge feature fusion map. The edge feature fusion function is: ； Where: Indicates the location Fusion edge features of pixels; Represents the infrared image temperature edge feature weight; Represents the visual edge feature weight of the RGB image; S33: The enhanced infrared image data is spliced ​​with the enhanced RGB image data to form four-channel image data.

6. The method for detecting eccentric load faults of a belt conveyor based on dual-source images according to claim 1, characterized in that: Step S40 specifically includes: S41: Divide the dual-source edge feature fusion image into a left and a right part using the center line as a dividing line, count the number of edge pixels in the left and right parts respectively, and calculate the percentage of the number of edge pixels in the left and right parts to the total number of edge pixels respectively; S42: According to the percentage of edge pixels in the left and right parts of the dual-source edge feature fusion image to the total edge pixels, set the overload fault alarm level, alarm triggering conditions and response measures of the belt conveyor (1): When the eccentric load fault alarm level is level 3, that is, yellow alarm, it is within the allowable eccentric load range. The alarm trigger condition is the difference in the percentage of edge pixels in the left and right parts of the dual-source edge feature fusion image to the total edge pixels. If the rate is greater than 4% and less than or equal to 10%, the response measure is to record the abnormality and increase the monitoring frequency; When the eccentric load fault alarm level is level 2, that is, orange alarm, it is a minor eccentric load fault. The alarm trigger condition is the difference in the percentage of edge pixels in the left and right parts of the dual-source edge feature fusion image to the total edge pixels. If the rate is greater than 10% and less than or equal to 20%, the response measure is to reduce speed and conduct manual inspection; When the eccentric load fault alarm level is level 1, that is, red alarm, it is a serious eccentric load fault. The alarm trigger condition is the difference in the percentage of edge pixels in the left and right parts of the dual-source edge feature fusion image to the total edge pixels. If it is greater than 20%, the response measure is to immediately shut down the machine for maintenance.

7. The method for detecting eccentric load faults of a belt conveyor based on dual-source images according to claim 2, characterized in that: The central processing module is a computer.

8. The method for detecting eccentric load faults of a belt conveyor based on dual-source images according to claim 2, characterized in that: The pressure sensor (5) is an electronic belt scale.

9. The method for detecting eccentric load faults of a belt conveyor based on dual-source images according to claim 2, characterized in that: The speed sensor (4) is an incremental rotary encoder.

10. The method for detecting eccentric load faults of a belt conveyor based on dual-source images according to claim 7, characterized in that: The belt conveyor (1), RGB camera (2), infrared camera (3), speed sensor (4), and pressure sensor (5) are all electrically connected to a computer.

Images