Conveyor belt deviation detection method, detection device and processor

Abstract

The present application provides a method, detection device, and processor for detecting conveyor belt deviation. The method includes: obtaining historical image data of the conveyor belt; constructing a neural network model; obtaining current image data of the conveyor belt; and using the neural network model to detect the current image data to determine whether the conveyor belt has deviated. In this solution, a new neural network model is constructed, which has higher computational efficiency. Using the new neural network model to detect the edge of the conveyor belt can adapt to complex environments, has stronger computational power and higher accuracy than traditional algorithms, and has better connectivity of the conveyor belt edge. Therefore, in actual environments, the edge of the conveyor belt can be stably detected, and the detected edge can accurately determine whether the conveyor belt has deviated, thereby improving the efficiency of detecting whether the conveyor belt has deviated.

Description

Technical Field

[0001] The present application relates to the field of image processing, and more specifically, to a method, a detection device, a computer-readable storage medium, and a processor for detecting conveyor belt deviation. Background Art

[0002] Conveyor belt deviation is one of the most common belt conveyor failures. It can cause material to spill and affect the safe operation of the conveyor equipment. Currently, the method used is to collect real-time video of the conveyor belts in the mine tunnels of coal preparation plants and monitor them remotely or conduct on-site inspections of the conveyor belts. Manual inspection is time-consuming and labor-intensive, and has low efficiency. Another method is to combine image recognition technology and edge detection algorithms to detect conveyor belt deviation. However, the current detection algorithms are slow and time-consuming, resulting in low detection efficiency. Summary of the Invention

[0003] The main purpose of this application is to provide a method, a detection device, a computer-readable storage medium and a processor for detecting whether a conveyor belt is deviating, so as to solve the problem of low efficiency in detecting whether a conveyor belt is deviating in the prior art.

[0004] According to one aspect of an embodiment of the present invention, a method for detecting whether a conveyor belt has deviated is provided, comprising: acquiring historical image data of the conveyor belt, wherein the historical image data is acquired by an image acquisition device installed above the conveyor belt; constructing a neural network model, wherein the neural network is trained using multiple sets of training data, each set of the multiple sets of training data including the historical image data and relevant information of the conveyor belt corresponding to the historical image data; acquiring current image data of the conveyor belt; and detecting the current image data using the neural network model to determine whether the conveyor belt has deviated.

[0005] Optionally, constructing a neural network model includes: obtaining target pixel points and non-target pixel points in the target area, wherein the target pixel point refers to the pixel point at the center of the target area, and the non-target pixel point refers to the pixel point in the target area other than the target pixel point, and the target area refers to the area including the edge of the conveyor belt in the historical image; determining a first area and a second area, wherein the first area refers to an area where the brightness difference between the target pixel point and the non-target pixel point is greater than or equal to a brightness threshold, and the second area refers to an area where the brightness difference between the target pixel point and the non-target pixel point is less than the brightness threshold; and respectively dividing the first area into the target pixel point and the non-target pixel point. The image data and the image data of the second area are operated with a kernel function to obtain first brightness data and second brightness data; the first brightness data and the second brightness data are input into a radial basis function neural network model of Gaussian projection for operation, wherein the radial basis function neural network model of Gaussian projection includes an input layer, an expansion layer, a hidden layer and an output layer, the input layer is used to receive input data, the expansion layer is used to subtract two adjacent input data, the hidden layer is used to calculate the input data after subtraction with a Gaussian function, and the output layer is used to output a calculation result, wherein the input data is the absolute value of the product of the first brightness data and the second brightness data.

[0006] Optionally, after constructing the neural network model, the method further includes: determining a center point, where the center point refers to the input data of the hidden layer of the neural network model; obtaining the maximum and minimum pixel values of the input data, and the number of pixel points of the input data; determining the variance based on the maximum pixel value, the minimum pixel value and the number of pixel points, and using an implicit function; and calculating the output weight of the neural network model using the least squares method, where the output weight refers to the weight value of the output data of the hidden layer in the calculation result of the output layer.

[0007] Optionally, after constructing the neural network model, the method further includes: obtaining standard image data; calculating an error value between a calculation result output by the neural network model and the standard image data; when the error value is less than an error threshold, determining that the neural network model has been trained; when the error value is greater than or equal to the error threshold, determining that the neural network model needs to be retrained, and retraining the neural network model.

[0008] Optionally, the neural network model is used to detect the current image data to determine whether the conveyor belt has deviated, including: using the neural network model to determine the first edge position and the second edge position of the current image data; obtaining the standard center line position of the conveyor belt, the standard center line position refers to the position of the center line of the conveyor belt when it has not deviated; and determining whether the conveyor belt has deviated based on the first edge position, the second edge position and the standard center line position.

[0009] Optionally, determining whether the conveyor belt has deviated based on the first edge position, the second edge position and the standard center line position includes: calculating in real time a first position difference between the first edge position and the standard center line position; calculating in real time a second position difference between the second edge position and the standard center line position; determining that the conveyor belt has deviated when the first position difference is less than or equal to a position threshold, or the second position difference is less than or equal to the position threshold; and determining that the conveyor belt has not deviated when the first position difference is greater than the position threshold and the second position difference is greater than the position threshold.

[0010] Optionally, when it is determined that the conveyor belt has deviated, the method further includes: generating an alarm message, wherein the alarm message is used to prompt that the conveyor belt has deviated; and controlling the conveyor belt to suspend operation.

[0011] According to another aspect of an embodiment of the present invention, a device for detecting conveyor belt deviation is provided, including: a first acquisition unit, used to acquire historical image data of the conveyor belt, wherein the historical image data is acquired by an image acquisition device, and the image acquisition device is installed above the conveyor belt; a construction unit, used to construct a neural network model, wherein the neural network is trained using multiple sets of training data, and each set of training data in the multiple sets of training data includes the historical image data and relevant information of the conveyor belt corresponding to the historical image data; a second acquisition unit, used to acquire current image data of the conveyor belt; and a detection unit, used to detect the current image data using the neural network model to determine whether the conveyor belt has deviated.

[0012] According to yet another aspect of the embodiments of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium includes a stored program, wherein the program executes any one of the methods described above.

[0013] According to yet another aspect of the embodiments of the present invention, a processor is provided, wherein the processor is configured to run a program, wherein any one of the methods is executed when the program is run.

[0014] In an embodiment of the present invention, historical image data of the conveyor belt is first acquired, followed by a neural network model. Current image data of the conveyor belt is then acquired, and finally, the neural network model is used to detect the current image data to determine whether the conveyor belt has deviated. This solution constructs a new neural network model with higher computational efficiency. Using this new neural network model to detect the edge of the conveyor belt can adapt to complex environments, offering greater computational power and accuracy than traditional algorithms. The connectivity of the conveyor belt edge is also improved, allowing for stable detection of the conveyor belt edge in practical environments. The detected edge can accurately determine whether the conveyor belt has deviated, thereby improving the efficiency of detecting whether the conveyor belt has deviated. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The drawings that constitute part of this application are used to provide a further understanding of this application. The illustrative embodiments of this application and their descriptions are used to explain this application and do not constitute an improper limitation on this application. In the drawings:

[0016] Figure 1 A schematic flow chart of a method for detecting conveyor belt deviation according to an embodiment of the present application is shown;

[0017] Figure 2 The structural diagram of the constructed neural network model is shown;

[0018] Figure 3 A schematic diagram showing the algorithm processing results;

[0019] Figure 4 A schematic structural diagram of a device for detecting conveyor belt deviation according to an embodiment of the present application is shown;

[0020] Figure 5 A flow chart of another method for detecting conveyor belt deviation according to an embodiment of the present application is shown.

[0021] The above drawings include the following reference numerals:

[0022] 100, conveyor belt; 200, first edge position; 300, second edge position; 400, standard center line position. DETAILED DESCRIPTION

[0023] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in this application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0024] In order to enable those skilled in the art to better understand the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of this application.

[0025] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present application described here. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0026] It should be understood that when an element (such as a layer, film, region, or substrate) is described as being "on" another element, the element may be directly on the other element or intervening elements may be present. Moreover, in the specification and claims, when it is described that an element is "connected to" another element, the element may be "directly connected to" the other element or "connected to" the other element through a third element.

[0027] Image edge detection is a major problem in image processing and computer vision. A large number of studies have shown that edge detection plays an important role in image high-order feature extraction, feature description, target recognition and image segmentation. How to quickly and accurately locate and extract image edge feature information has become one of the research hotspots. Edge detection algorithms are relatively rich. By integrating new theories, they can be roughly divided into gradient-based edge detection algorithms, mathematical morphology-based edge detection algorithms, genetic algorithm-based edge detection algorithms, and neural network-based edge detection algorithms.

[0028] For real-time processing of working videos of conveyor belts in coal preparation plant mines, we need algorithms that can adapt to relatively complex environments. On the other hand, we also need algorithms that consume low computing resources and can achieve real-time processing when receiving real-time monitoring videos.

[0029] Because image edges contain a large amount of background information and important structural information, traditional edge detection methods often prioritize handcrafted low-level features such as color, brightness, and gradients. These methods, such as the Sobel operator, Sparse Code Gradients (SCG), and Structured Forest Edge Detection (SE) algorithms, are often used for edge detection. While edge detection methods using low-level features have made significant progress, their limitations are also evident. With the development of deep learning technology, particularly the emergence of convolutional neural networks (CNNs), it has been discovered that neural networks possess powerful automatic learning capabilities in the field of natural images. However, CNN-based edge detection methods rely on classification networks. Due to the slowness of classification methods, detection takes a long time, is inflexible, and suffers from high noise levels, resulting in low detection efficiency.

[0030] As mentioned in the background technology, the efficiency of detecting whether a conveyor belt is deviating in the prior art is low. In order to solve the above problem, in a typical embodiment of the present application, a method for detecting whether a conveyor belt is deviating, a detection device, a computer-readable storage medium and a processor are provided.

[0031] According to an embodiment of the present application, a method for detecting conveyor belt deviation is provided.

[0032] Figure 1 Flowchart of the method for detecting conveyor belt deviation according to an embodiment of the present application. Figure 1 As shown, the method includes the following steps:

[0033] Step S101, acquiring historical image data of a conveyor belt, wherein the historical image data is acquired by an image acquisition device installed above the conveyor belt;

[0034] Specifically, the image acquisition device may be a camera, a video camera, a scanner, etc.

[0035] Step S102: constructing a neural network model, wherein the neural network is trained using multiple sets of training data, each set of training data including the historical image data and information related to the conveyor belt corresponding to the historical image data;

[0036] Step S103, obtaining current image data of the conveyor belt;

[0037] Step S104: Use the neural network model to detect the current image data to determine whether the conveyor belt is deviating.

[0038] In the above method, historical image data of the conveyor belt is first acquired, followed by a neural network model. Current image data of the conveyor belt is then acquired, and finally the neural network model is used to detect the current image data to determine whether the conveyor belt has deviated. This solution constructs a new neural network model with higher computational efficiency. Using this new neural network model to detect the edge of the conveyor belt can adapt to complex environments, offering greater computational power and accuracy than traditional algorithms. The connectivity of the conveyor belt edge is also improved, allowing for stable detection of the conveyor belt edge in real-world environments. The detected edge can accurately determine whether the conveyor belt has deviated, thereby improving the efficiency of detecting whether the conveyor belt has deviated.

[0039] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0040] In one embodiment, the neural network model can be a radial basis neural network model of Gaussian projection. Specifically, the image acquisition device can sample the image data of the conveyor belt at predetermined intervals. The collected historical images are preprocessed. Since the environment of the coal preparation plant is relatively dark and the image data has many noise points, a Gaussian filter can be used to smooth the image. The Gaussian filter calculates the weighted average value of the original image, that is, the value of each pixel point can be obtained by weighted averaging the original value of the pixel point and the value of the adjacent pixel point. During the calculation, each pixel point is traversed in turn, and the pixel point calculated in each round is taken as the center point (0, 0). Considering the σ×σ pixels in the adjacent squares, the weight value of each pixel point is determined by the Gaussian function:

[0041]

[0042] Here, x represents the horizontal coordinate of the pixel relative to the center point, y represents the vertical coordinate of the pixel relative to the center point, σ represents the width of the Gaussian function, and G(x,y) represents the Gaussian function of the pixel, that is, the weight of that position during the Gaussian filtering process. Based on the weight of each pixel, a weighted average is taken for the σ×σ pixels within the width. The final value obtained is the pixel grayscale value after Gaussian filtering. When all pixels have been calculated, the image is Gaussian filtered, completing the preprocessing. Each pixel in the preprocessed image is then input into a radial basis function neural network based on Gaussian projection.

[0043] In one embodiment of the present application, a neural network model is constructed, including: obtaining target pixel points and non-target pixel points in a target area, the target pixel point refers to the pixel point at the center of the target area, the non-target pixel point refers to the pixel point other than the target pixel point in the target area, and the target area refers to the area including the edge of the conveyor belt in the historical image; determining a first area and a second area, the first area refers to an area where the brightness difference between the target pixel point and the non-target pixel point is greater than or equal to a brightness threshold, and the second area refers to an area where the brightness difference between the target pixel point and the non-target pixel point is less than the brightness threshold; respectively The image data of one area and the image data of the second area are calculated with a kernel function to obtain first brightness data and second brightness data; the first brightness data and the second brightness data are input into a radial basis neural network model of Gaussian projection for calculation, wherein the radial basis neural network model of Gaussian projection includes an input layer, an expansion layer, a hidden layer and an output layer, the input layer is used to receive input data, the expansion layer is used to subtract two adjacent input data, the hidden layer is used to calculate the input data after subtraction with a Gaussian function, and the output layer is used to output the calculation result, wherein the input data is the absolute value of the product of the first brightness data and the second brightness data. In this embodiment, the radial basis neural network model of Gaussian projection has higher computational efficiency and better connectivity of the calculated edges, which can more efficiently and accurately analyze the edge information of the conveyor belt, thereby further improving the detection efficiency.

[0044] When the brightness changes dramatically, the human eye experiences a sense of edge enhancement, where the bright side appears brighter and the dark side appears darker. This characteristic is called the Mach effect. Using the Mach effect, the following two mathematical expressions can be proposed:

[0045] (1) Each pixel is compared with its surrounding points by a weighted distance, namely:

[0046]

[0047] (2) When comparing a certain pixel point, determine whether the surrounding points have a higher or lower pixel value than the central point, that is:

[0048]

[0049]

[0050] Wherein, D represents the entire image domain (target area), D1 represents the first area where the difference between the target pixel point x and the non-target pixel point y is greater than zero, D2 represents the second area where the difference between the pixel point x and the non-target pixel point y is less than zero, I(·) represents the pixel value of the point, K(·) is the Gaussian kernel function, Mid(·) represents the difference between the target pixel point x and the non-target pixel point y, GP (Gaussian Positive) and GN (Gaussian Negative) are the positive and negative parts of the difference between the target pixel point x and the non-target pixel point y, respectively, that is, the first brightness data and the second brightness data.

[0051] Based on the above, we can get the GP and GN values, and further perform the following steps to achieve edge contour highlighting:

[0052] (1) Let n = 1, take a pixel point xn and its 3 × 3 region in an image with h pixels, calculate the difference between the pixel value of xn and the non-target pixel point, and obtain regions D1 and D2;

[0053] (2) Multiply the values of the two regions by the kernel function Kernel and add them together to obtain the GP and GN values respectively;

[0054] (3) The value of xn is reassigned to -GN×GP;

[0055] (4) If n<h, then n is increased by 1, and the next pixel is taken and the process is repeated until n=h.

[0056] The above process can be understood as the process of Gaussian projection. A Gaussian projection radial basis neural network model is further constructed to achieve the purpose of highlighting the contour. The structure diagram of the model is as follows Figure 2 As shown in the figure, the input layer consists of a pixel xn in an image with h pixels. The difference between xn and the pixel value of the non-target pixel is calculated, and the result is input to the second expansion layer. The third layer is the hidden layer. The hidden layer uses the Gaussian function as the implicit function. The implicit function is: Among them, x is the output result of the previous layer, the parameter represents the variance of the Gaussian function, and the weight values of all nodes in the hidden layer are the same. In the Gaussian projection radial basis neural network model, the weights of all nodes except the hidden layer are 1.

[0057] In another embodiment of the present application, after constructing the neural network model, the method further includes: determining the center point, the center point refers to the input data of the hidden layer of the neural network model; obtaining the maximum and minimum pixel values of the input data, and the number of pixel points of the input data; determining the variance based on the maximum value of the pixel value, the minimum value of the pixel value and the number of pixel points, and using an implicit function; and calculating the output weight of the neural network model using the least squares method, the output weight refers to the weight value of the output data of the hidden layer in the calculation result of the output layer. In this solution, after constructing the neural network model, the neural network must be further trained to ensure that the computational efficiency of the trained neural network model is higher.

[0058] Specifically, the nodes of the hidden layer are d i , i = 1, 2, ... h, h is the total number of input image data, let the corresponding center point be each input data, the input data can be directly mapped to the hidden layer.

[0059] Specifically, the number of nodes in the hidden layer is the same as the number of input data. In order not to lose image information, according to the above formula, the center point is set to each input data. This not only ensures the integrity of the image information, but also further improves the efficiency of the model.

[0060] The neural network model uses the Gaussian function as the implicit function, and the variance calculation formula is: σ represents variance, Xmax represents the maximum pixel value, Xmin represents the minimum pixel value, and h represents the number of pixels. This direct way of solving the variance can obtain a more suitable variance for different images, which speeds up the training efficiency to a certain extent.

[0061] Since the number of nodes and variance in the neural network model are already determined, the input to output can be represented by a linear equation, and its output weight can be solved by the least squares method. The formula is as follows: i=1,2,...,h,p=1,2,...,P, where ω represents the output weight, represents the pth input image, p = 1, 2, ..., P, P is the total number of input images, d i Represents each center point, that is, each node of the hidden layer.

[0062] The Gaussian Projection RBF Neural Network utilizes the sparse connections and weight sharing characteristics of CNNs, reducing the complexity of the neural network model. Images with multidimensional input vectors can be directly modeled, reducing the complexity of feature extraction. Therefore, this neural network model can be considered an extension of the RBF Neural Network, adding fixed weights to the first layer. The filtered image then serves as the input sample for training the Gaussian Projection RBF Neural Network.

[0063] In another embodiment of the present application, after constructing the neural network model, the method further includes: obtaining standard image data; calculating the error value between the calculation result output by the neural network model and the standard image data; when the error value is less than the error threshold, determining that the training of the neural network model has been completed; when the error value is greater than or equal to the error threshold, determining that the neural network model needs to be retrained, and retraining the neural network model. In this embodiment, by detecting the error value calculated by the neural network model and retraining the neural network model when the error value is large, the efficiency of the neural network model can be further ensured.

[0064] In another embodiment of the present application, the neural network model is used to detect the current image data to determine whether the conveyor belt has deviated, including: determining a first edge position and a second edge position of the current image data using the neural network model; obtaining a standard centerline position of the conveyor belt, wherein the standard centerline position refers to the position of the centerline of the conveyor belt when it is not deviating; and determining whether the conveyor belt has deviated based on the first edge position, the second edge position, and the standard centerline position. In this embodiment, by statistically analyzing the first and second edge positions of the processed image, and then based on the first edge position, the second edge position, and the standard centerline position, it is possible to further accurately determine whether the conveyor belt has deviated.

[0065] In a specific embodiment of the present application, determining whether the conveyor belt has deviated based on the first edge position, the second edge position, and the standard centerline position includes: calculating a first position difference between the first edge position and the standard centerline position in real time; calculating a second position difference between the second edge position and the standard centerline position in real time; determining that the conveyor belt has deviated if the first position difference is less than or equal to a position threshold, or if the second position difference is less than or equal to the position threshold; and determining that the conveyor belt has not deviated if the first position difference is greater than the position threshold and the second position difference is greater than the position threshold. In this embodiment, if the difference between the first edge position or the second edge position of the conveyor belt and the standard centerline position is greater than the position threshold, it indicates that the conveyor belt may not have deviated at this time, or the degree of deviation is not significant and has no significant impact on the conveyor belt. If the difference between the first edge position or the second edge position of the conveyor belt and the standard centerline position is less than or equal to the position threshold, it indicates that the conveyor belt has deviated at this time and the edge of the conveyor belt may have approached the standard centerline position. Therefore, this embodiment can further accurately determine whether the conveyor belt has deviated.

[0066] In practical applications, the edge contour can be accurately determined by collecting the current image data and then using a neural network model. Based on the output results, a contour tracking algorithm can be used to extract the edge to achieve the purpose of edge detection. If the edge position in 5 consecutive frames of images exceeds the median position, it proves that the transmission belt has deviated, and subsequent early warning or alarm processing is required.

[0067] Specifically, the schematic diagram of the algorithm processing result is as follows Figure 3 As shown, the camera sensor is installed above the conveyor belt 100 in the coal preparation plant mine, and the camera direction is along the working direction of the conveyor belt 100. The result of algorithm processing can determine the first edge position 200 and the second edge position 300 of the conveyor belt 100. It can be determined that the conveyor belt 100 is currently working normally, and the first edge position 200 and the second edge position 300 have not reached the standard center line position 400. If the first edge position 200 and the second edge position 300 reach the standard center line position 400, a deviation warning is issued.

[0068] In another specific embodiment of the present application, upon determining that the conveyor belt has deviated, the method further comprises: generating an alarm message, the alarm message being used to indicate that the conveyor belt has deviated; and controlling the conveyor belt to suspend operation. In this embodiment, by generating an alarm message, a worker can be promptly notified that the conveyor belt has deviated, and then the conveyor belt can be controlled to suspend operation, thereby avoiding dangerous accidents caused by the conveyor belt deviating.

[0069] Image edge detection plays a vital role in fields such as image processing and computer vision. When processing images of mine conveyor belts, the algorithm in this solution incorporates the characteristics of the human visual system and has the ability to adapt to relatively complex environments, a significant advantage over traditional edge detection algorithms. Furthermore, compared to the classic CNN convolutional neural network edge detection method, the algorithm in this solution consumes fewer computing resources and has a shorter detection time. It also achieves better edge detection results when processing images containing a lot of Gaussian noise. This is a significant advantage in the relatively dark and noisy environment of coal preparation plants, enabling stable detection of conveyor belt edges.

[0070] Furthermore, based on the edge data obtained for each frame, the distance between the conveyor belt edge pixel and the image's standard centerline can be detected to quickly determine whether the conveyor belt is operating correctly. Furthermore, the deviation warning function improves system stability by accumulating errors across multiple frames.

[0071] The present application also provides a device for detecting conveyor belt deviation. It should be noted that the device for detecting conveyor belt deviation in the present application can be used to implement the method for detecting conveyor belt deviation provided in the present application. The following describes the device for detecting conveyor belt deviation provided in the present application.

[0072] Figure 4 Schematic diagram of a device for detecting conveyor belt deviation according to an embodiment of the present application. Figure 4 As shown, the device includes:

[0073] A first acquisition unit 10 is configured to acquire historical image data of the conveyor belt, wherein the historical image data is acquired by an image acquisition device installed above the conveyor belt;

[0074] Specifically, the image acquisition device may be a camera, a video camera, a scanner, etc.

[0075] A construction unit 20 is configured to construct a neural network model, wherein the neural network is trained using multiple sets of training data, each set of training data including the historical image data and information related to the conveyor belt corresponding to the historical image data;

[0076] A second acquiring unit 30 is used to acquire current image data of the conveyor belt;

[0077] The detection unit 40 is used to detect the current image data using the neural network model to determine whether the conveyor belt is deviating.

[0078] In the above-mentioned scheme, the first acquisition unit acquires historical image data of the conveyor belt, the construction unit constructs a neural network model, the second acquisition unit acquires current image data of the conveyor belt, and the detection unit uses the neural network model to detect the current image data to determine whether the conveyor belt has deviated. In this scheme, a new neural network model is constructed with higher computational efficiency. Using this new neural network model to detect the edge of the conveyor belt can adapt to complex environments, has stronger computational power and higher accuracy than traditional algorithms, and has better connectivity of the conveyor belt edge. Therefore, in actual environments, the edge of the conveyor belt can be stably detected, and the detected edge can accurately determine whether the conveyor belt has deviated, thereby improving the efficiency of detecting whether the conveyor belt has deviated.

[0079] In one embodiment, the neural network model can be a radial basis neural network model of Gaussian projection. Specifically, the image acquisition device can sample the image data of the conveyor belt at predetermined intervals. The collected historical images are preprocessed. Since the environment of the coal preparation plant is relatively dark and the image data has many noise points, a Gaussian filter can be used to smooth the image. The Gaussian filter calculates the weighted average value of the original image, that is, the value of each pixel point can be obtained by weighted averaging the original value of the pixel point and the value of the adjacent pixel point. During the calculation, each pixel point is traversed in turn, and the pixel point calculated in each round is taken as the center point (0, 0). Considering the σ×σ pixels in the adjacent squares, the weight value of each pixel point is determined by the Gaussian function:

[0080]

[0081] Here, x represents the horizontal coordinate of the pixel relative to the center point, y represents the vertical coordinate of the pixel relative to the center point, σ represents the width of the Gaussian function, and G(x,y) represents the Gaussian function of the pixel, that is, the weight of that position during the Gaussian filtering process. Based on the weight of each pixel, a weighted average is taken for the σ×σ pixels within the width. The final value obtained is the pixel grayscale value after Gaussian filtering. When all pixels have been calculated, the image is Gaussian filtered, completing the preprocessing. Each pixel in the preprocessed image is then input into a radial basis function neural network based on Gaussian projection.

[0082] In one embodiment of the present application, the construction unit includes a first acquisition module, a first determination module, a calculation module and a processing module, the first acquisition module is used to obtain target pixel points and non-target pixel points in the target area, the above-mentioned target pixel point refers to the pixel point in the center of the above-mentioned target area, the above-mentioned non-target pixel point refers to the pixel point other than the above-mentioned target pixel point in the above-mentioned target area, and the above-mentioned target area refers to the area including the edge of the above-mentioned conveyor belt in the historical image; the first determination module is used to determine the first area and the second area, the above-mentioned first area refers to the area where the brightness difference between the above-mentioned target pixel point and the above-mentioned non-target pixel point is greater than or equal to the brightness threshold, and the above-mentioned second area refers to the area where the above-mentioned brightness difference between the above-mentioned target pixel point and the above-mentioned non-target pixel point is less than the above-mentioned brightness threshold The area of the value; the operation module is used to respectively operate the image data of the first area and the image data of the second area with the kernel function to obtain the first brightness data and the second brightness data; the processing module is used to input the first brightness data and the second brightness data into the radial basis neural network model of Gaussian projection for operation, wherein the radial basis neural network model of Gaussian projection includes an input layer, an expansion layer, a hidden layer and an output layer, the input layer is used to receive input data, the expansion layer is used to subtract two adjacent input data, the hidden layer is used to calculate the input data after subtraction with the Gaussian function, and the output layer is used to output the calculation result, wherein the input data is the absolute value of the product of the first brightness data and the second brightness data. In this embodiment, the radial basis neural network model of Gaussian projection has higher computational efficiency and better connectivity of the calculated edges, which can more efficiently and accurately analyze the edge information of the conveyor belt, thereby further improving the detection efficiency.

[0083] When the brightness changes dramatically, the human eye experiences a sense of edge enhancement, where the bright side appears brighter and the dark side appears darker. This characteristic is called the Mach effect. Using the Mach effect, the following two mathematical expressions can be proposed:

[0084] (1) Each pixel is compared with its surrounding points by a weighted distance, namely:

[0085]

[0086] (2) When comparing a certain pixel point, determine whether the surrounding points have a higher or lower pixel value than the central point, that is:

[0087]

[0088]

[0089] Wherein, D represents the entire image domain (target area), D1 represents the first area where the difference between the target pixel point x and the non-target pixel point y is greater than zero, D2 represents the second area where the difference between the pixel point x and the non-target pixel point y is less than zero, I(·) represents the pixel value of the point, K(·) is the Gaussian kernel function, Mid(·) represents the difference between the target pixel point x and the non-target pixel point y, GP (Gaussian Positive) and GN (Gaussian Negative) are the positive and negative parts of the difference between the target pixel point x and the non-target pixel point y, respectively, that is, the first brightness data and the second brightness data.

[0090] Based on the above, we can get the GP and GN values, and further perform the following steps to achieve edge contour highlighting:

[0091] (1) Let n = 1, take a pixel point xn and its 3 × 3 region in an image with h pixels, calculate the difference between the pixel value of xn and the non-target pixel point, and obtain regions D1 and D2;

[0092] (2) Multiply the values of the two regions by the kernel function Kernel and add them together to obtain the GP and GN values respectively;

[0093] (3) The value of xn is reassigned to -GN×GP;

[0094] (4) If n<h, then n is increased by 1, and the next pixel is taken and the process is repeated until n=h.

[0095] The above process can be understood as the process of Gaussian projection. A Gaussian projection radial basis neural network model is further constructed to achieve the purpose of highlighting the contour. The structure diagram of the model is as follows Figure 2 As shown in the figure, the input layer consists of a pixel xn in an image with h pixels. The difference between xn and the pixel value of the non-target pixel is calculated, and the result is input to the second expansion layer. The third layer is the hidden layer. The hidden layer uses the Gaussian function as the implicit function. The implicit function is: Among them, x is the output result of the previous layer, the parameter represents the variance of the Gaussian function, and the weight values of all nodes in the hidden layer are the same. In the Gaussian projection radial basis neural network model, the weights of all nodes except the hidden layer are 1.

[0096] In another embodiment of the present application, the above-mentioned device also includes a first determination unit, a third acquisition unit, a second determination unit and a first calculation unit. The first determination unit is used to determine the center point after constructing the neural network model. The above-mentioned center point refers to the above-mentioned input data of the above-mentioned hidden layer of the above-mentioned neural network model; the third acquisition unit is used to obtain the maximum and minimum values of the pixel values of the above-mentioned input data, as well as the number of pixel points of the above-mentioned input data; the second determination unit is used to determine the variance based on the maximum value of the above-mentioned pixel value, the minimum value of the above-mentioned pixel value and the number of the above-mentioned pixel points, and using an implicit function; the first calculation unit is used to calculate the output weight of the above-mentioned neural network model using the least squares method. The above-mentioned output weight refers to the weight value of the output data of the above-mentioned hidden layer in the above-mentioned calculation result of the above-mentioned output layer. In this solution, after the neural network model is constructed, the neural network must be further trained to ensure that the computational efficiency of the trained neural network model is higher.

[0097] Specifically, the nodes of the hidden layer are d i , i = 1, 2, ... h, h is the total number of input image data, let the corresponding center point be each input data, the input data can be directly mapped to the hidden layer.

[0098] Specifically, the number of nodes in the hidden layer is the same as the number of input data. In order not to lose image information, according to the above formula, the center point is set to each input data. This not only ensures the integrity of the image information, but also further improves the efficiency of the model.

[0099] The neural network model uses the Gaussian function as the implicit function, and the variance calculation formula is: σ represents variance, Xmax represents the maximum pixel value, Xmin represents the minimum pixel value, and h represents the number of pixels. This direct way of solving the variance can obtain a more suitable variance for different images, which speeds up the training efficiency to a certain extent.

[0100] Since the number of nodes and variance in the neural network model are already determined, the input to output can be represented by a linear equation, and its output weight can be solved by the least squares method. The formula is as follows: i=1,2,...,h,p=1,2,...,P, where ω represents the output weight, represents the p-th input image, p = 1, 2, ..., P, P is the total number of input images, and d represents each center point, that is, each node of the hidden layer.

[0101] The Gaussian Projection RBF Neural Network utilizes the sparse connections and weight sharing characteristics of CNNs, reducing the complexity of the neural network model. Images with multidimensional input vectors can be directly modeled, reducing the complexity of feature extraction. Therefore, this neural network model can be considered an extension of the RBF Neural Network, adding fixed weights to the first layer. The filtered image then serves as the input sample for training the Gaussian Projection RBF Neural Network.

[0102] In another embodiment of the present application, the above-mentioned device further includes a fourth acquisition unit, a second calculation unit, a third determination unit and a training unit, wherein the fourth acquisition unit is used to acquire standard image data after constructing the neural network model; the second calculation unit is used to calculate the error value between the calculation result output by the above-mentioned neural network model and the above-mentioned standard image data; the third determination unit is used to determine that the above-mentioned neural network model has been trained when the above-mentioned error value is less than the error threshold; the training unit is used to determine that the above-mentioned neural network model needs to be retrained when the above-mentioned error value is greater than or equal to the above-mentioned error threshold, and retrain the above-mentioned neural network model. In this embodiment, by detecting the error value calculated by the neural network model and retraining the neural network model when the error value is large, the efficiency of the neural network model can be further ensured.

[0103] In another embodiment of the present application, the detection unit includes a second determination module, a second acquisition module, and a third determination module. The second determination module is configured to use the neural network model to determine the first edge position and the second edge position of the current image data. The second acquisition module is configured to obtain the standard centerline position of the conveyor belt, where the standard centerline position refers to the position of the centerline of the conveyor belt when it is not deviating. The third determination module is configured to determine whether the conveyor belt has deviated based on the first edge position, the second edge position, and the standard centerline position. In this embodiment, by statistically analyzing the first and second edge positions of the processed image, and then based on the first edge position, the second edge position, and the standard centerline position, it is possible to further accurately determine whether the conveyor belt has deviated.

[0104] In a specific embodiment of the present application, the third determination module includes a first calculation submodule, a second calculation submodule, a first determination submodule and a second determination submodule, the first calculation submodule is used to calculate the first position difference between the above-mentioned first edge position and the above-mentioned standard center line position in real time; the second calculation submodule is used to calculate the second position difference between the above-mentioned second edge position and the above-mentioned standard center line position in real time; the first determination submodule is used to determine that the above-mentioned conveyor belt has deviated when the above-mentioned first position difference is less than or equal to the position threshold, or the above-mentioned second position difference is less than or equal to the above-mentioned position threshold; the second determination submodule is used to determine that the above-mentioned conveyor belt has not deviated when the above-mentioned first position difference is greater than the above-mentioned position threshold, and the above-mentioned second position difference is greater than the above-mentioned position threshold. In this embodiment, when the difference between the first edge position or the second edge position of the conveyor belt and the standard center line position is greater than the position threshold, it indicates that the conveyor belt may not have deviated at this time, or the degree of deviation is not large, and there is not much impact on the conveyor belt. If the difference between the first edge position or the second edge position and the standard center line position is less than or equal to the position threshold, it indicates that the conveyor belt has deviated at this time, and the edge of the conveyor belt may have approached the standard center line position. Therefore, this embodiment can further accurately determine whether the conveyor belt has deviated.

[0105] In practical applications, the edge contour can be accurately determined by collecting the current image data and then using a neural network model. Based on the output results, a contour tracking algorithm can be used to extract the edge to achieve the purpose of edge detection. If the edge position in 5 consecutive frames of images exceeds the median position, it proves that the transmission belt has deviated, and subsequent early warning or alarm processing is required.

[0106] Specifically, the schematic diagram of the algorithm processing result is as follows Figure 3 As shown, the camera sensor is installed above the conveyor belt 100 in the coal preparation plant mine, and the camera direction is along the working direction of the conveyor belt 100. The result of algorithm processing can determine the first edge position 200 and the second edge position 300 of the conveyor belt 100. It can be determined that the conveyor belt 100 is currently working normally, and the first edge position 200 and the second edge position 300 have not reached the standard center line position 400. If the first edge position 200 and the second edge position 300 reach the standard center line position 400, a deviation warning is issued.

[0107] In another specific embodiment of the present application, the device further includes a generating unit and a controlling unit. The generating unit is configured to generate an alarm message upon determining that the conveyor belt has deviated, the alarm message being used to indicate that the conveyor belt has deviated; and the controlling unit is configured to control the conveyor belt to suspend operation. In this embodiment, by generating the alarm message, a worker can be promptly notified that the conveyor belt has deviated, and the conveyor belt can then be controlled to suspend operation, thereby avoiding dangerous accidents caused by the conveyor belt deviating.

[0108] Image edge detection plays a vital role in fields such as image processing and computer vision. When processing images of mine conveyor belts, the algorithm in this solution incorporates the characteristics of the human visual system and has the ability to adapt to relatively complex environments, a significant advantage over traditional edge detection algorithms. Furthermore, compared to the classic CNN convolutional neural network edge detection method, the algorithm in this solution consumes fewer computing resources and has a shorter detection time. It also achieves better edge detection results when processing images containing a lot of Gaussian noise. This is a significant advantage in the relatively dark and noisy environment of coal preparation plants, enabling stable detection of conveyor belt edges.

[0109] Furthermore, based on the edge data obtained for each frame, the distance between the conveyor belt edge pixel and the image's standard centerline can be detected to quickly determine whether the conveyor belt is operating correctly. Furthermore, the deviation warning function improves system stability by accumulating errors across multiple frames.

[0110] The above-mentioned conveyor belt deviation detection device includes a processor and a memory. The above-mentioned first acquisition unit, construction unit, second acquisition unit and detection unit are all stored in the memory as program units, and the processor executes the above-mentioned program units stored in the memory to realize the corresponding functions.

[0111] The processor contains a kernel, which retrieves the corresponding program unit from the memory. One or more kernels can be set, and the efficiency of detecting whether the conveyor belt is running off can be improved by adjusting the kernel parameters.

[0112] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0113] An embodiment of the present invention provides a computer-readable storage medium having a program stored thereon, which implements the above-mentioned method for detecting conveyor belt deviation when the program is executed by a processor.

[0114] An embodiment of the present invention provides a processor, which is used to run a program, wherein the method for detecting conveyor belt deviation is executed when the program is run.

[0115] An embodiment of the present invention provides a device, comprising a processor, a memory, and a program stored in the memory and executable on the processor. When the processor executes the program, at least the following steps are performed:

[0116] Step S101, acquiring historical image data of a conveyor belt, wherein the historical image data is acquired by an image acquisition device installed above the conveyor belt;

[0117] Step S102: constructing a neural network model, wherein the neural network is trained using multiple sets of training data, each set of training data including the historical image data and information related to the conveyor belt corresponding to the historical image data;

[0118] Step S103, obtaining current image data of the conveyor belt;

[0119] Step S104: Use the neural network model to detect the current image data to determine whether the conveyor belt is deviating.

[0120] The devices in this article can be servers, PCs, PADs, mobile phones, etc.

[0121] The present application also provides a computer program product, which, when executed on a data processing device, is adapted to execute a program for initializing at least the following method steps:

[0122] Step S101, acquiring historical image data of a conveyor belt, wherein the historical image data is acquired by an image acquisition device installed above the conveyor belt;

[0123] Step S102: constructing a neural network model, wherein the neural network is trained using multiple sets of training data, each set of training data including the historical image data and information related to the conveyor belt corresponding to the historical image data;

[0124] Step S103, obtaining current image data of the conveyor belt;

[0125] Step S104: Use the neural network model to detect the current image data to determine whether the conveyor belt is deviating.

[0126] In order to enable those skilled in the art to more clearly understand the technical solution of the present application, the technical solution and technical effects of the present application will be explained below with reference to specific embodiments.

[0127] Example

[0128] This embodiment relates to a method for detecting conveyor belt deviation. Figure 5 As shown,

[0129] First, start testing;

[0130] The camera sensor collects images, which are historical image data;

[0131] Obtain training samples and perform Gaussian filtering on the training samples;

[0132] Training the Gaussian projection radial basis neural network model includes: determining the center point, calculating the variance of the implicit function, and updating the network weights;

[0133] Determine whether the error value of the neural network model is less than the error threshold;

[0134] When the error value is greater than or equal to the error threshold, the neural network model is retrained;

[0135] When the error value is less than the error threshold, it is determined that the neural network model has been trained;

[0136] The camera sensor collects images, and what is collected at this time is the current image data;

[0137] Obtaining a working sample and performing Gaussian filtering on the working sample;

[0138] Use the trained Gaussian projection radial basis neural network model to detect the current image;

[0139] Determine whether the edge position of the conveyor belt coincides with the standard center line position;

[0140] In the case of overlap, a conveyor belt deviation warning is generated and the conveyor belt is controlled to stop working;

[0141] In the case of non-overlap, make sure the conveyor belt is working properly;

[0142] End the detection process.

[0143] In the above embodiments of the present invention, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0144] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only exemplary. For example, the division of the above-mentioned units can be a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0145] The units described above as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0146] In addition, the functional units in the various embodiments of the present invention may be integrated into a single processing unit, each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0147] If the above-mentioned integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server or network device, etc.) to execute all or part of the steps of the above-mentioned methods of each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk.

[0148] From the above description, it can be seen that the above embodiments of the present application achieve the following technical effects:

[0149] 1) The conveyor belt deviation detection method of the present application first obtains historical image data of the conveyor belt, then constructs a neural network model, then obtains current image data of the conveyor belt, and finally uses the neural network model to detect the current image data to determine whether the conveyor belt has deviated. In this solution, a new neural network model is constructed, which has higher computational efficiency. Using the new neural network model to detect the edge of the conveyor belt can adapt to complex environments, has stronger computational power and higher accuracy than traditional algorithms, and has better connectivity of the conveyor belt edge. Therefore, in actual environments, the edge of the transmission belt can be stably detected, and the detected edge can accurately determine whether the conveyor belt has deviated, thereby improving the efficiency of detecting whether the conveyor belt has deviated.

[0150] 2) The conveyor belt deviation detection device of the present application comprises a first acquisition unit that acquires historical image data of the conveyor belt, a construction unit that constructs a neural network model, a second acquisition unit that acquires current image data of the conveyor belt, and a detection unit that uses the neural network model to detect the current image data to determine whether the conveyor belt has deviated. In this solution, a new neural network model is constructed, which has higher computational efficiency. Using the new neural network model to detect the edge of the conveyor belt can adapt to complex environments, has stronger computational power and higher accuracy than traditional algorithms, and has better connectivity of the conveyor belt edge. Furthermore, in actual environments, the edge of the conveyor belt can be stably detected, and the detected edge can be used to accurately determine whether the conveyor belt has deviated, thereby improving the efficiency of detecting whether the conveyor belt has deviated.

[0151] The above description is merely a preferred embodiment of the present application and is not intended to limit the present application. Various modifications and variations are possible for those skilled in the art. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present application shall be included within the scope of protection of the present application.

Claims

1. A method for detecting conveyor belt deviation, characterized in that: include: Acquiring historical image data of the conveyor belt, wherein the historical image data is acquired by an image acquisition device installed above the conveyor belt; Constructing a neural network model, wherein the neural network is trained using multiple sets of training data, each set of training data including the historical image data and information related to the conveyor belt corresponding to the historical image data; Acquiring current image data of the conveyor belt; The neural network model is used to detect the current image data to determine whether the conveyor belt is deviating. Constructing a neural network model includes: obtaining target pixels and non-target pixels in a target area, wherein the target pixel refers to a pixel at the center of the target area, and the non-target pixel refers to a pixel other than the target pixel in the target area, and the target area refers to an area including the edge of the conveyor belt in the historical image; determining a first area and a second area, wherein the first area refers to an area where the brightness difference between the target pixel and the non-target pixel is greater than or equal to a brightness threshold, and the second area refers to an area where the brightness difference between the target pixel and the non-target pixel is less than the brightness threshold; and respectively converting the image of the first area into a plurality of pixels. The data and the image data of the second area are operated with a kernel function to obtain first brightness data and second brightness data; the first brightness data and the second brightness data are input into a radial basis neural network model of Gaussian projection for operation, wherein the radial basis neural network model of Gaussian projection includes an input layer, an expansion layer, a hidden layer and an output layer, the input layer is used to receive input data, the expansion layer is used to differ between two adjacent input data, the hidden layer is used to calculate the input data after difference with a Gaussian function, and the output layer is used to output a calculation result, wherein the input data is the absolute value of the product of the first brightness data and the second brightness data.

2. The method according to claim 1, characterized in that After constructing the neural network model, the method further includes: Determining a center point, wherein the center point refers to the input data of the hidden layer of the neural network model; Obtaining the maximum and minimum pixel values of the input data, as well as the number of pixels of the input data; Determining the variance using an implicit function according to the maximum value of the pixel value, the minimum value of the pixel value, and the number of pixels; The least squares method is used to calculate the output weight of the neural network model, where the output weight refers to the weight value of the output data of the hidden layer in the calculation result of the output layer.

3. The method according to claim 1, characterized in that After constructing the neural network model, the method further includes: Obtain standard image data; Calculating an error value between a calculation result output by the neural network model and the standard image data; When the error value is less than the error threshold, determining that the training of the neural network model has been completed; When the error value is greater than or equal to the error threshold, it is determined that the neural network model needs to be retrained, and the neural network model is retrained.

4. The method according to claim 1, wherein Using the neural network model to detect the current image data to determine whether the conveyor belt is deviating includes: Determining a first edge position and a second edge position of the current image data using the neural network model; Obtaining a standard centerline position of the conveyor belt, wherein the standard centerline position refers to the centerline position of the conveyor belt when it is not deviated; Whether the conveyor belt is deviated is determined according to the first edge position, the second edge position and the standard center line position.

5. The method according to claim 4, characterized in that Determining whether the conveyor belt is deviating according to the first edge position, the second edge position, and the standard centerline position includes: Calculating in real time a first position difference between the first edge position and the standard center line position; Calculating in real time a second position difference between the second edge position and the standard center line position; When the first position difference is less than or equal to a position threshold, or the second position difference is less than or equal to the position threshold, determining that the conveyor belt has deviated; When the first position difference is greater than the position threshold and the second position difference is greater than the position threshold, it is determined that the conveyor belt is not deviating.

6. The method according to any one of claims 1 to 5, characterized in that In the case where it is determined that the conveyor belt has deviated, the method further includes: Generate an alarm message, wherein the alarm message is used to prompt that the conveyor belt has deviated; Control the conveyor belt to stop working.

7. A device for detecting deviation of a conveyor belt, characterized in that: include: A first acquisition unit is configured to acquire historical image data of the conveyor belt, wherein the historical image data is acquired by an image acquisition device installed above the conveyor belt; a construction unit configured to construct a neural network model, wherein the neural network is trained using a plurality of sets of training data, each set of the plurality of training data including the historical image data and information related to the conveyor belt corresponding to the historical image data; a second acquiring unit, configured to acquire current image data of the conveyor belt; A detection unit is used to detect the current image data using the neural network model to determine whether the conveyor belt is deviating. The construction unit includes a first acquisition module, a first determination module, a calculation module and a processing module, wherein the first acquisition module is used to acquire target pixels and non-target pixels in a target area, wherein the target pixel refers to a pixel at the center of the target area, and the non-target pixel refers to a pixel other than the target pixel in the target area, and the target area refers to an area including the edge of the conveyor belt in the historical image; the first determination module is used to determine a first area and a second area, wherein the first area refers to an area where the brightness difference between the target pixel and the non-target pixel is greater than or equal to a brightness threshold, and the second area refers to an area where the brightness difference between the target pixel and the non-target pixel is less than the brightness threshold; The operation module is used to respectively operate the image data of the first area and the image data of the second area with the kernel function to obtain first brightness data and second brightness data; the processing module is used to input the first brightness data and the second brightness data into a Gaussian projection radial basis neural network model for operation, wherein the Gaussian projection radial basis neural network model includes an input layer, an expansion layer, a hidden layer and an output layer, the input layer is used to receive input data, the expansion layer is used to subtract two adjacent input data, the hidden layer is used to calculate the input data after subtraction with the Gaussian function, and the output layer is used to output the calculation result, wherein the input data is the absolute value of the product of the first brightness data and the second brightness data.

8. A computer-readable storage medium, characterized in that The computer-readable storage medium includes a stored program, wherein the program executes the method according to any one of claims 1 to 6.

9. A processor, characterized in that: The processor is configured to run a program, wherein the program executes the method according to any one of claims 1 to 6 when running.

Images