Coal gangue grading recognition system based on deep learning

Abstract

The invention belongs to the field of coal gangue grading recognition, and particularly discloses a coal gangue grading recognition system based on deep learning, which comprises a coal gangue image acquisition module, a coal gangue edge separation module and a coal gangue classification module. According to the scheme, gray level inversion is carried out on the initial image to improve the detail definition, a bounded sum operator is applied to realize migration of low-brightness area pixels to a high-brightness area, adaptive threshold processing is adopted to screen edge pixels, and complete and accurate extraction of coal gangue edges is realized; the method comprises the following steps: extracting texture features by adopting empirical wavelet transform, fusing mean value, contrast and homogeneity features calculated by a gray-level co-occurrence matrix, building a deep convolutional pulse neural network fused with a pulse coding technology, introducing an arithmetic optimization algorithm to optimize key parameters of a coal gangue classification model, capturing spatial features and time sequence pulse features of a coal gangue image, and constructing a deep convolutional pulse neural network fused with a pulse coding technology; and comprehensive mining and efficient fusion of the multidimensional key features of the coal gangue image are realized.

Description

Technical Field

[0001] This invention relates to the field of coal gangue grading and identification, specifically to a coal gangue grading and identification system based on deep learning. Background Technology

[0002] Coal gangue is a solid waste generated during coal mining and washing. Its composition and characteristics vary, and deep learning models can automatically learn the complex features of coal gangue, resulting in higher accuracy and faster processing efficiency. However, in the initial X-ray images of coal gangue during mining and washing, low-brightness areas often exhibit blurred details, insufficient contrast, and edges with similar grayscale values ​​to the background, leading to edge breakage, false edge interference, or loss of detail. This results in inaccurate basic data for subsequent classification and limited overall recognition accuracy, failing to meet the precise separation requirements of industrial production. Traditional coal gangue classification models cannot comprehensively capture multi-dimensional information such as spatial structure and frequency distribution in X-ray images. Their feature fusion methods are simple and lack the ability to uncover deep features, leading to poor adaptability and insufficient generalization ability for coal gangue samples with complex compositions and diverse morphologies. Furthermore, the identification boundaries between different grades of coal gangue are blurred, easily resulting in misclassification or omission. Summary of the Invention

[0003] To address the above issues and overcome the shortcomings of existing technologies, this invention provides a deep learning-based coal gangue grading and identification system. Addressing the technical problem that initial X-ray images of coal gangue during coal mining and washing are prone to edge breakage, false edge interference, or detail loss in low-brightness areas, failing to meet the precise separation requirements of industrial production, this solution performs grayscale inversion on the initial image to improve detail clarity. It applies bounded sum operators to migrate pixels from low-brightness areas to high-brightness areas and employs adaptive thresholding to filter edge pixels, completely overcoming the limitations of traditional fixed thresholds. This system can accurately adapt to the image features of coal gangue with different compositions and morphologies, achieving complete and accurate extraction of coal gangue edges. Furthermore, it addresses the issue that traditional coal gangue classification models use simple feature fusion methods, making it difficult to mine deep features, resulting in poor adaptability and insufficient generalization ability for different coal gangue samples. The identification of graded coal gangue often suffers from blurred boundaries, leading to misclassification and omissions. This solution addresses this issue by fusing multi-dimensional features. Empirical wavelet transform is used to extract texture features, and one-dimensional Fourier transform is employed to obtain frequency domain features. Based on these frequency domain features, row and column filter banks are constructed. The mean, contrast, and homogeneity features calculated from the gray-level co-occurrence matrix are then fused to form a comprehensive texture and frequency domain fusion feature set. Simultaneously, a deep convolutional pulse neural network incorporating pulse coding technology is built, and arithmetic optimization algorithms are introduced to optimize key parameters of the coal gangue classification model. This captures the spatial and temporal pulse features of coal gangue images, enabling comprehensive mining and efficient fusion of multi-dimensional key features. This clarifies the identification boundaries of different grades of coal gangue, maintaining high recognition accuracy even in low-quality images and rare sample scenarios. Furthermore, it significantly improves the model training convergence speed and shortens the training cycle.

[0004] The technical solution adopted by the present invention is as follows: The present invention provides a coal gangue grading and identification system based on deep learning, the coal gangue grading and identification system based on deep learning includes a coal gangue image acquisition module, a coal gangue edge separation module and a coal gangue classification module;

[0005] The coal gangue image acquisition module uses X-rays to acquire initial images of the coal gangue;

[0006] The coal gangue edge separation module uses an image processing method based on soft computation operators to accurately extract the coal gangue edges from the initial coal gangue image, thereby obtaining the coal gangue edge image.

[0007] The coal gangue classification module uses a classification and recognition method to construct a coal gangue classification model and uses an algorithm optimization operator to optimize the key parameters in the coal gangue classification model, and performs hierarchical recognition of coal gangue edge images.

[0008] Furthermore, in the coal gangue edge separation module, the image processing method based on soft computing operators specifically includes the following steps:

[0009] Step A1: Perform grayscale inversion on the initial coal gangue image to improve detail clarity. Apply a bounded sum operator to the inverted initial coal gangue image, denoted as the bounded image, to achieve pixel migration from low-brightness areas to high-brightness areas. The formula used is as follows: ;

[0010] In the formula, It is a bounded image. These are the row and column coordinates of the image pixels. This indicates that the output bounded image pixel grayscale values ​​do not exceed 255. This is the inverted initial image of the coal gangue;

[0011] Step A2: Based on the first and second derivatives of the image in the horizontal and vertical directions, apply sharpening filtering to the bounded image to eliminate image blur and obtain a sharpened image;

[0012] Step A3: Calculate the gradient magnitude. Calculate the gradient magnitude of the sharpened image to quantify the intensity of the coal gangue edges, obtaining the gradient image. The formula for calculating the gradient magnitude is as follows: ;

[0013] In the formula, It is the gradient magnitude. It is the first derivative in the horizontal direction. It is the first derivative in the vertical direction;

[0014] Step A4: Adaptive thresholding. Apply adaptive thresholding to the gradient image to filter out coal gangue edge pixels, calculate the average gray value of the gradient image, and set the adaptive threshold according to the range of the average gray value to obtain the binary edge image, i.e., the coal gangue edge image.

[0015] Furthermore, in the coal gangue classification module, the classification and identification method includes a coal gangue classification model construction method and a coal gangue classification model optimization method. The coal gangue classification model construction method specifically includes the following steps:

[0016] Step B1: Image acquisition, acquiring a coal gangue classification image set, which includes coal gangue X-ray images and corresponding coal gangue classification levels;

[0017] Step B2: Feature extraction. The texture features of the X-ray image of coal gangue are extracted using empirical wavelet transform. The empirical wavelet transform parameters are initialized by setting the number of rows and columns of the empirical wavelet transform.

[0018] Step B3: One-dimensional Fourier transform. Perform one-dimensional Fourier transform on each row and column of the coal gangue X-ray image to obtain the frequency domain features.

[0019] Step B4: Filter bank construction. Based on the frequency domain features, row and column filter banks are constructed to obtain the gray-level co-occurrence matrix. Based on the gray-level co-occurrence matrix, the mean, contrast features, and homogeneity features of the image are calculated. The calculation formula for the homogeneity features is as follows: ;

[0020] In the formula, It is a homogeneous characteristic. and These are the row and column coordinates of the image pixels. and They are respectively for and Traversal, and It is the center pixel value of a local region. It is a gray-level co-occurrence matrix;

[0021] Step B5: Feature set output. The mean extracted from the row filter bank and column filter bank is fused with the contrast feature and homogeneity feature to obtain the sub-band feature image, forming the final texture feature set.

[0022] Step B6: Establish and initialize a deep convolutional spiking neural network as a coal gangue classification model. The deep convolutional spiking neural network integrates convolutional feature extraction and pulse coding technology to capture the spatial features of coal gangue X-ray images. It includes 2 convolutional layers, 2 max pooling layers, 2 dropout layers, 1 pulse coding layer, 1 flattening layer and 3 fully connected layers.

[0023] The convolutional layer calculates the convolutional feature map of the final texture feature set through the convolutional filter. The max pooling layer performs non-overlapping downsampling on the convolutional feature map to obtain a pooling feature map with reduced dimensionality. The dropout layer masks some neurons during training and restores all neurons during testing, outputting a regularized feature map. The pulse coding layer converts the feature vector into a pulse sequence and calculates the membrane potential change to obtain a multidimensional pulse sequence. The flattening layer converts the multidimensional pulse sequence into a one-dimensional feature vector. The fully connected layer calculates the classification probability of coal gangue. The formula for calculating the membrane potential change is as follows: ;

[0024] In the formula, yes Membrane potential at time t, It is a membrane resistor. It is the input current. It is a film capacitor.

[0025] Furthermore, the coal gangue classification model optimization method introduces an arithmetic optimization algorithm to optimize the key parameters in the deep convolutional spiking neural network, specifically including the following steps:

[0026] Step C1: Parameter initialization, determine the optimization parameter set as the weights and biases of the convolutional layer and the fully connected layer, the refractory period of the pulse coding layer and the membrane time constant, and randomly generate candidate parameter combinations as the initial population;

[0027] Step C2: Fitness function calculation. The classification accuracy of the deep convolutional spiking neural network on the coal gangue classification image set is used as the fitness function. The fitness value of each candidate parameter combination is calculated, and the current optimal parameter combination is selected.

[0028] Step C3: Parameter update. Calculate the mathematical optimization probability of the current optimal parameter combination. Randomly generate two comparison thresholds with values ​​in the range [0, 1], denoted as the first threshold and the second threshold, respectively. If the mathematical optimization probability of the current optimal parameter combination is less than the first threshold, update the current optimal parameter combination. If the second threshold is less than 0.5, use the division operator to update the current optimal parameter combination. If the second threshold is greater than or equal to 0.5, use the multiplication operator to update the current optimal parameter combination.

[0029] Step C4: Parameter refinement. If the first threshold is less than or equal to the mathematical optimization probability of the current optimal parameter combination, a third threshold with a value range of [0, 1] is randomly generated. If the third threshold is less than 0.5, the current optimal parameter combination is updated using a subtraction operator. If the third threshold is greater than or equal to 0.5, the current optimal parameter combination is updated using an addition operator.

[0030] Step C5: After iteration is complete, repeat steps C2 to C4 until the iteration is complete to obtain the globally optimal parameter combination. Input the globally optimal parameter combination into the deep convolutional spiking neural network, use the cross-loss entropy function for training, and output the final classification result of coal gangue.

[0031] The beneficial effects achieved by adopting the above solution are as follows:

[0032] (1) In view of the technical problem that the low brightness area of ​​the initial X-ray image of coal gangue in the coal mining and washing process is prone to edge breakage, false edge interference or loss of details, which cannot meet the requirements of precise separation in industrial production, this solution performs gray-level inversion on the initial image to improve the clarity of details, applies bounded sum operators to realize the migration of pixels in low brightness area to high brightness area, and uses adaptive threshold processing to filter edge pixels, completely getting rid of the limitations of traditional fixed thresholds, and can accurately adapt to the image features of coal gangue with different components and different forms, realizing the complete and accurate extraction of coal gangue edges;

[0033] (2) In view of the technical problems of traditional coal gangue classification model feature fusion method is simple and difficult to explore deep features, resulting in poor adaptability and insufficient generalization ability of the model to coal gangue samples, blurred identification boundary of coal gangue of different grades, and easy misclassification and omission, this scheme uses multi-dimensional feature fusion, uses empirical wavelet transform to extract texture features, uses one-dimensional Fourier transform to obtain frequency domain features, constructs row and column filter banks based on frequency domain features, and fuses the mean, contrast and homogeneity features calculated by gray-level co-occurrence matrix to form a comprehensive texture and frequency domain fusion feature set. At the same time, a deep convolutional pulse neural network with pulse coding technology is built, and an arithmetic optimization algorithm is introduced to optimize the key parameters of coal gangue classification model, capture the spatial features and temporal pulse features of coal gangue image, realize the comprehensive mining and efficient fusion of multi-dimensional key features of coal gangue image, make the identification boundary of coal gangue of different grades clear, and maintain high recognition accuracy in low quality image and rare sample scenarios. At the same time, it greatly improves the model training convergence speed and shortens the training cycle. Attached Figure Description

[0034] Figure 1 A module connection diagram of a deep learning-based coal gangue grading and identification system provided by the present invention;

[0035] Figure 2 This is a flowchart of the steps of a classification and identification method provided by the present invention.

[0036] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof. Detailed Implementation

[0037] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0038] Example 1, see Figure 1 The present invention provides a coal gangue grading and identification system based on deep learning. The coal gangue grading and identification system based on deep learning includes a coal gangue image acquisition module, a coal gangue edge separation module, and a coal gangue classification module.

[0039] The coal gangue image acquisition module uses X-rays to acquire initial images of the coal gangue;

[0040] The coal gangue edge separation module uses an image processing method based on soft computation operators to accurately extract the coal gangue edges from the initial coal gangue image, thereby obtaining the coal gangue edge image.

[0041] The coal gangue classification module uses a classification and recognition method to construct a coal gangue classification model and uses an algorithm optimization operator to optimize the key parameters in the coal gangue classification model, and performs hierarchical recognition of coal gangue edge images.

[0042] Example 2, see Figure 1 and Figure 2 This embodiment, based on the above embodiment, describes an image processing method based on soft computation operators in the coal gangue edge separation module, specifically including the following steps:

[0043] Step A1: Perform grayscale inversion on the initial coal gangue image to improve detail clarity. Apply a bounded sum operator to the inverted initial coal gangue image, denoted as the bounded image, to achieve pixel migration from low-brightness areas to high-brightness areas. The formula used is as follows: ;

[0044] In the formula, It is a bounded image. These are the row and column coordinates of the image pixels. This indicates that the output bounded image pixel grayscale values ​​do not exceed 255. This is the inverted initial image of the coal gangue;

[0045] Step A2: Based on the first and second derivatives of the image in the horizontal and vertical directions, apply sharpening filtering to the bounded image to eliminate image blur and obtain a sharpened image;

[0046] Step A3: Calculate the gradient magnitude. Calculate the gradient magnitude of the sharpened image to quantify the intensity of the coal gangue edges, obtaining the gradient image. The formula for calculating the gradient magnitude is as follows: ;

[0047] In the formula, It is the gradient magnitude. It is the first derivative in the horizontal direction. It is the first derivative in the vertical direction;

[0048] Step A4: Adaptive thresholding. Apply adaptive thresholding to the gradient image to filter out coal gangue edge pixels, calculate the average gray value of the gradient image, and set the adaptive threshold according to the range of the average gray value to obtain the binary edge image, i.e., the coal gangue edge image.

[0049] Example 3, see Figure 1 and Figure 2 This embodiment is based on the above embodiment. In the coal gangue classification module, the classification and identification method includes a coal gangue classification model construction method and a coal gangue classification model optimization method. The coal gangue classification model construction method specifically includes the following steps:

[0050] Step B1: Image acquisition, acquiring a coal gangue classification image set, which includes coal gangue X-ray images and corresponding coal gangue classification levels;

[0051] Step B2: Feature extraction. The texture features of the X-ray image of coal gangue are extracted using empirical wavelet transform. The empirical wavelet transform parameters are initialized by setting the number of rows and columns of the empirical wavelet transform.

[0052] Step B3: One-dimensional Fourier transform. Perform one-dimensional Fourier transform on each row and column of the coal gangue X-ray image to obtain the frequency domain features.

[0053] Step B4: Filter bank construction. Based on the frequency domain features, row and column filter banks are constructed to obtain the gray-level co-occurrence matrix. Based on the gray-level co-occurrence matrix, the mean, contrast features, and homogeneity features of the image are calculated. The calculation formula for the homogeneity features is as follows: ;

[0054] In the formula, It is a homogeneous characteristic. and These are the row and column coordinates of the image pixels. and They are respectively for and Traversal, and It is the center pixel value of a local region. It is a gray-level co-occurrence matrix;

[0055] Step B5: Feature set output. The mean extracted from the row filter bank and column filter bank is fused with the contrast feature and homogeneity feature to obtain the sub-band feature image, forming the final texture feature set.

[0056] Step B6: Establish and initialize a deep convolutional spiking neural network as a coal gangue classification model. The deep convolutional spiking neural network integrates convolutional feature extraction and pulse coding technology to capture the spatial features of coal gangue X-ray images. It includes 2 convolutional layers, 2 max pooling layers, 2 dropout layers, 1 pulse coding layer, 1 flattening layer and 3 fully connected layers.

[0057] The convolutional layer calculates the convolutional feature map of the final texture feature set through the convolutional filter. The max pooling layer performs non-overlapping downsampling on the convolutional feature map to obtain a pooling feature map with reduced dimensionality. The dropout layer masks some neurons during training and restores all neurons during testing, outputting a regularized feature map. The pulse coding layer converts the feature vector into a pulse sequence and calculates the membrane potential change to obtain a multidimensional pulse sequence. The flattening layer converts the multidimensional pulse sequence into a one-dimensional feature vector. The fully connected layer calculates the classification probability of coal gangue. The formula for calculating the membrane potential change is as follows: ;

[0058] In the formula, yes Membrane potential at time t, It is a membrane resistor. It is the input current. It is a film capacitor.

[0059] Example 4, see Figure 1 and Figure 2 This embodiment is based on the above embodiment. The coal gangue classification model optimization method introduces an arithmetic optimization algorithm to optimize the key parameters in the deep convolutional spiking neural network, specifically including the following steps:

[0060] Step C1: Parameter initialization, determine the optimization parameter set as the weights and biases of the convolutional layer and the fully connected layer, the refractory period of the pulse coding layer and the membrane time constant, and randomly generate candidate parameter combinations as the initial population;

[0061] Step C2: Fitness function calculation. The classification accuracy of the deep convolutional spiking neural network on the coal gangue classification image set is used as the fitness function. The fitness value of each candidate parameter combination is calculated, and the current optimal parameter combination is selected.

[0062] Step C3: Parameter update. Calculate the mathematical optimization probability of the current optimal parameter combination. Randomly generate two comparison thresholds with values ​​in the range [0, 1], denoted as the first threshold and the second threshold, respectively. If the mathematical optimization probability of the current optimal parameter combination is less than the first threshold, update the current optimal parameter combination. If the second threshold is less than 0.5, use the division operator to update the current optimal parameter combination. If the second threshold is greater than or equal to 0.5, use the multiplication operator to update the current optimal parameter combination.

[0063] Step C4: Parameter refinement. If the first threshold is less than or equal to the mathematical optimization probability of the current optimal parameter combination, a third threshold with a value range of [0, 1] is randomly generated. If the third threshold is less than 0.5, the current optimal parameter combination is updated using a subtraction operator. If the third threshold is greater than or equal to 0.5, the current optimal parameter combination is updated using an addition operator.

[0064] Step C5: After iteration is complete, repeat steps C2 to C4 until the iteration is complete to obtain the globally optimal parameter combination. Input the globally optimal parameter combination into the deep convolutional spiking neural network, use the cross-loss entropy function for training, and output the final classification result of coal gangue.

[0065] Example 5 is based on the above examples. In Example 2, adaptive thresholding is used to calculate the average gray value of the gradient image and set an adaptive threshold based on the range of the average gray value. The adaptive thresholding rules are as follows: ;

[0066] In the formula, It is a binary edge image, specifically a coal gangue edge image. It is the average gray value. It is a gradient image.

[0067] Example 6 is based on the above examples. In Example 3, the feature extraction uses empirical wavelet transform to extract the texture features of the X-ray image of coal gangue. The empirical wavelet transform parameters are initialized by setting the number of rows and columns of the empirical wavelet transform. The number of rows and columns is the same as the size of the coal gangue edge image.

[0068] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0069] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

[0070] The present invention and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present invention; the actual structure is not limited thereto. In conclusion, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the invention, such designs should fall within the protection scope of the present invention.

Claims

1. A deep learning-based coal gangue grading and identification system, characterized in that, It includes a coal gangue image acquisition module, a coal gangue edge separation module, and a coal gangue classification module; The coal gangue image acquisition module uses X-rays to acquire initial images of the coal gangue; The coal gangue edge separation module uses an image processing method based on soft computation operators to accurately extract the coal gangue edges from the initial coal gangue image, thereby obtaining the coal gangue edge image. The coal gangue classification module uses a classification and recognition method to construct a coal gangue classification model and uses an algorithm optimization operator to optimize the key parameters in the coal gangue classification model, and performs hierarchical recognition of coal gangue edge images.

2. The deep learning-based coal gangue grading and identification system according to claim 1, characterized in that, In the coal gangue edge separation module, the image processing method based on soft computing operators specifically includes the following steps: Step A1: Perform grayscale inversion processing on the initial coal gangue image, apply the bounded sum operator to the inverted initial coal gangue image, and denote it as the bounded image; Step A2: Based on the first and second derivatives of the image in the horizontal and vertical directions, apply a sharpening filter to the bounded image to obtain a sharpened image; Step A3: Calculate the gradient magnitude. Calculate the gradient magnitude of the sharpened image to quantify the intensity of the coal gangue edges and obtain the gradient image. Step A4: Adaptive thresholding. Apply adaptive thresholding to the gradient image to filter out coal gangue edge pixels, calculate the average gray value of the gradient image, and set the adaptive threshold according to the range of the average gray value to obtain the coal gangue edge image.

3. The deep learning-based coal gangue grading and identification system according to claim 2, characterized in that, In the coal gangue classification module, the classification and identification method includes a coal gangue classification model construction method and a coal gangue classification model optimization method. The coal gangue classification model construction method specifically includes the following steps: Step B1: Image acquisition, acquiring a coal gangue classification image set, which includes coal gangue X-ray images and corresponding coal gangue classification levels; Step B2: Feature extraction. The texture features of the X-ray image of coal gangue are extracted using empirical wavelet transform. The empirical wavelet transform parameters are initialized by setting the number of rows and columns of the empirical wavelet transform. Step B3: One-dimensional Fourier transform. Perform one-dimensional Fourier transform on each row and column of the coal gangue X-ray image to obtain the frequency domain features. Step B4: Filter bank construction. Based on the frequency domain features, construct row filter banks and column filter banks to obtain the gray-level co-occurrence matrix. Calculate the mean, contrast features, and homogeneity features of the image based on the gray-level co-occurrence matrix. Step B5: Feature set output. The mean extracted from the row filter bank and column filter bank is fused with the contrast feature and homogeneity feature to obtain the sub-band feature image, forming the final texture feature set. Step B6: Establish and initialize a deep convolutional spiking neural network as a coal gangue classification model. The deep convolutional spiking neural network integrates convolutional feature extraction and pulse coding technology to capture the spatial features of coal gangue X-ray images, including convolutional layers, max pooling layers, dropout layers, pulse coding layers, flattening layers, and fully connected layers. The convolutional layer calculates the convolutional feature map of the final texture feature set through the convolutional filter. The max pooling layer performs non-overlapping downsampling on the convolutional feature map to obtain a pooling feature map with reduced dimensionality. The dropout layer masks some neurons during training and restores all neurons during testing, outputting a regularized feature map. The pulse coding layer converts the feature vector into a pulse sequence and calculates the membrane potential change to obtain a multidimensional pulse sequence. The flattening layer converts the multidimensional pulse sequence into a one-dimensional feature vector. The fully connected layer calculates the classification probability of coal gangue.

4. The deep learning-based coal gangue grading and identification system according to claim 3, characterized in that, The coal gangue classification model optimization method introduces an arithmetic optimization algorithm to optimize key parameters in a deep convolutional spiking neural network, specifically including the following steps: Step C1: Parameter initialization, determine the optimization parameter set as the weights and biases of the convolutional and fully connected layers, the refractory period of the pulse coding layer and the membrane time constant, and randomly generate candidate parameter combinations as the initial population; Step C2: Fitness function calculation. The classification accuracy of the deep convolutional spiking neural network on the coal gangue classification image set is used as the fitness function. The fitness value of each candidate parameter combination is calculated, and the current optimal parameter combination is selected. Step C3: Parameter update. Calculate the mathematical optimization probability of the current optimal parameter combination. Randomly generate two comparison thresholds with values ​​in the range [0, 1], denoted as the first threshold and the second threshold, respectively. If the mathematical optimization probability of the current optimal parameter combination is less than the first threshold, update the current optimal parameter combination. If the second threshold is less than 0.5, use the division operator to update the current optimal parameter combination. If the second threshold is greater than or equal to 0.5, use the multiplication operator to update the current optimal parameter combination. Step C4: Parameter refinement. If the first threshold is less than or equal to the mathematical optimization probability of the current optimal parameter combination, a third threshold with a value range of [0, 1] is randomly generated. If the third threshold is less than 0.5, the current optimal parameter combination is updated using a subtraction operator. If the third threshold is greater than or equal to 0.5, the current optimal parameter combination is updated using an addition operator. Step C5: After iteration is complete, repeat steps C2 to C4 until the iteration is complete to obtain the globally optimal parameter combination. Input the globally optimal parameter combination into the deep convolutional spiking neural network, use the cross-loss entropy function for training, and output the final classification result of coal gangue.

Images