Coal and rock identification method and system based on image and XRF spectrum information fusion

Abstract

The present disclosure provides a coal-rock identification method and system based on the fusion of image and XRF spectrum information, which relates to the technical field of geological rock identification and classification, and includes: obtaining a coal-rock image to be detected; inputting the coal-rock image into an image detection model to obtain suspicious coal gangue and its position information; performing element testing on the suspicious coal gangue to obtain element information, inputting the element information and the suspicious coal gangue image into a fusion classification model, and then extracting element features through an element feature extraction module, extracting image features through an image feature extraction module, performing matrix multiplication of image features and element features in an initial fusion module to form a similarity matrix, and then deeply fusion the similarity matrix with image features to obtain initial fusion features, and using the initial fusion features to obtain coal / coal gangue classification results.

Description

Technical Field

[0001] The present disclosure relates to the technical field of geological rock identification and classification, and in particular to a coal rock identification method and system based on the fusion of image and XRF spectrum information. Background Art

[0002] The statements in this section merely provide background information related to the present disclosure and do not necessarily constitute prior art.

[0003] At present, comprehensive mechanized coal mining (fully mechanized mining) has become the dominant mining method in coal production. In order to promote the development of coal mining in a safer and more efficient direction, less-manned or even unmanned coal mining has become an inevitable trend in the future. However, the distribution of coal seams faced by fully mechanized mining faces is complex and changeable, the hardness and thickness of coal rocks are unevenly distributed, and they are often affected by geological structures, resulting in fractures and folds in the occurrence of coal seams, which undoubtedly increases the difficulty of coal mining machines to accurately cut coal. Therefore, on-site coal and rock identification technology is particularly important for improving its ability to accurately cut coal.

[0004] The commonly used method for coal and rock identification in the early days was the natural gamma ray method. Currently, the coal and rock identification methods mainly involve infrared thermal imaging recognition technology, image recognition technology, spectrum recognition technology, radar detection technology, vibration recognition technology, etc.

[0005] However, studies have shown that the X-ray method has certain hazards and is difficult to promote and apply when the content of radioactive elements is low. In addition, the coal mining working face environment is harsh, and factors such as high noise, high dust concentration and high humidity will have varying degrees of impact on infrared, spectral, vibration and radar detection methods. The identification effect is not ideal and it is difficult to popularize and apply. Summary of the invention

[0006] In order to solve the above problems, the present invention proposes a coal rock identification method and system based on the fusion of image and XRF spectrum information, which fuses image features and element features, and jointly uses image-based recognition methods and fusion recognition methods to improve the system recognition efficiency and achieve accurate identification of coal gangue.

[0007] According to some embodiments, the present disclosure adopts the following technical solutions:

[0008] The coal and rock identification method based on the fusion of image and XRF spectrum information includes:

[0009] Acquire coal and rock images to be detected;

[0010] Inputting the coal-rock image into an image detection model to obtain suspicious gangue and its location information;

[0011] XRF spectrum element test is carried out on suspicious gangue to obtain element information, and the element information and the suspicious gangue image are input into the fusion classification model. The element features are extracted by the element feature extraction module, and the image features are extracted by the image feature extraction module. The image features and element features are matrix multiplied in the preliminary fusion module to form a similarity matrix. The similarity matrix and image features are deeply fused to obtain the preliminary fusion features, and the classification results of coal and gangue are obtained using the preliminary fusion features.

[0012] According to some embodiments, the present disclosure adopts the following technical solutions:

[0013] The coal and rock identification system based on the fusion of image and XRF spectrum information includes:

[0014] A data acquisition module, used to acquire the coal and rock images to be detected;

[0015] A preliminary detection module, used to input the coal-rock image into an image detection model to obtain suspicious coal gangue and its location information;

[0016] The deep identification module is used to perform XRF spectrum element testing on suspicious coal gangue to obtain element information, and input the element information and the suspicious coal gangue image into the fusion classification model. The fusion classification model extracts element features through the element feature extraction module and the image feature extraction module respectively, and the image features and element features are matrix multiplied in the initial fusion module to form a similarity matrix. The similarity matrix and the image features are deeply fused to obtain the initial fusion features, and the classification results of coal and coal gangue are obtained using the initial fusion features.

[0017] According to some embodiments, the present disclosure adopts the following technical solutions:

[0018] A computer program product includes a computer program, which, when executed by a processor, implements a coal rock identification method based on the fusion of image and XRF spectrum information.

[0019] According to some embodiments, the present disclosure adopts the following technical solutions:

[0020] A non-transitory computer-readable storage medium is used to store computer instructions. When the computer instructions are executed by a processor, a coal rock identification method based on the fusion of image and XRF spectrum information is implemented.

[0021] According to some embodiments, the present disclosure adopts the following technical solutions:

[0022] An electronic device comprises: a processor, a memory and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory so that the electronic device executes a coal rock identification method based on the fusion of image and XRF spectrum information.

[0023] Compared with the prior art, the present invention has the following beneficial effects:

[0024] The present invention discloses a coal rock identification method based on the fusion of image and XRF spectrum information, proposes a fusion identification method, takes image features and element features as input at the same time, realizes accurate identification of coal gangue, and solves the problems of dust pollution, uneven lighting, etc. that affect the accuracy of pure image-based identification methods; jointly uses image-based recognition methods and fusion recognition methods to construct a recognition model to improve model recognition efficiency.

[0025] The present invention discloses a coal-rock identification method based on the fusion of image and XRF spectrum information, and constructs an intelligent sorting system, which has the advantages of high equipment automation, fast identification speed, high lithology identification accuracy, etc., and has good application prospects; at the same time, it reduces labor, saves costs, and reduces the occurrence of occupational diseases such as pneumoconiosis.

[0026] The present invention discloses a coal and rock identification method based on the fusion of image and XRF spectrum information. Due to the different distribution characteristics of the two types of data, the splicing feature fusion in the traditional method is likely to cause feature loss and the relevant feature extraction module is difficult to train. The similarity matrix proposed in the present invention can measure the similarity of the two types of features, which can help the module adjust its focus without causing feature loss. The method is simple to implement and has good results.

[0027] The present invention discloses a coal rock identification method based on the fusion of image and XRF spectrum information, and proposes a coal rock feature extraction module, namely, a long-short distance feature extraction block. Since coal rock has significant self-similarity, that is, the local and the whole are very similar, coal rock has no obvious shape features, only color and texture, while conventional identification tasks have color, texture and shape features. Therefore, compared with conventional identification tasks, coal rock feature extraction is more difficult, and local feature extraction is more important. The long-short distance feature extraction block proposed in the present invention jointly uses 1×7, 7×1, and 3×3 deep separation convolution kernels to extract features. Compared with traditional convolution kernels, the proposed module has stronger regional feature extraction capabilities and can effectively improve the classification effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] The accompanying drawings constituting a part of the present disclosure are used to provide a further understanding of the present disclosure. The illustrative embodiments of the present disclosure and their descriptions are used to explain the present disclosure and do not constitute an improper limitation on the present disclosure.

[0029] Figure 1 This is a flow chart of a coal-rock identification method based on image and XRF spectrum information fusion according to an embodiment of the present disclosure;

[0030] Figure 2 It is a schematic diagram of the fusion process of an embodiment of the present disclosure;

[0031] Figure 3 An intelligent coal gangue sorting system according to an embodiment of the present disclosure;

[0032] Figure 4 Schematic diagram of a basic component module (one-dimensional residual convolution module) of Fcs according to an embodiment of the present disclosure;

[0033] Figure 5 Schematic diagram of a long- and short-distance feature joint extraction block according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0034] The present disclosure is further described below in conjunction with the accompanying drawings and embodiments.

[0035] It should be noted that the following detailed descriptions are all illustrative and are intended to provide further explanation of the present disclosure. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present disclosure belongs.

[0036] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present disclosure. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0037] Example 1

[0038] In one embodiment of the present disclosure, a method for identifying coal and rock based on the fusion of image and XRF spectrum information is provided, comprising:

[0039] Step 1: Obtain the coal rock image to be detected;

[0040] Step 2: input the coal rock image into the image detection model to obtain suspicious coal gangue and its location information;

[0041] Step three: perform elemental testing on the suspected gangue to obtain elemental information, and input the elemental information and the suspected gangue image into the fusion classification model. The fusion classification model extracts elemental features through the element feature extraction module and the image feature extraction module respectively. The image features and elemental features are matrix multiplied in the initial fusion module to form a similarity matrix. The similarity matrix is ​​deeply fused with the image features to obtain the initial fusion features, and the classification results of coal and gangue are obtained using the initial fusion features.

[0042] As an embodiment, the specific implementation process of a coal rock identification method based on image and XRF spectrum information fusion disclosed in the present invention is as follows:

[0043] Step 1: Obtain an image of the object to be inspected;

[0044] Using a camera installed directly above the conveyor belt, the camera collects images of coal and coal gangue on the conveyor belt in real time.

[0045] Step 2: Input the coal-rock image into a fusion classification model to obtain suspicious gangue and its location information;

[0046] The image detection model is a traditional YOLOv5 model, which calls the image detection model to detect objects in the image, determine whether there is suspicious coal gangue and location information, and output the classification result of the suspicious coal gangue. If there is suspicious coal gangue, the result also includes the coal gangue image and location information.

[0047] The training process of the image detection model is as follows: ① Collect images of coal gangue on the conveyor belt and annotate the coal gangue and coal in the images; ② Use the annotated data to create training sets, validation sets, and test sets; ③ Use the training sets, validation sets, and test sets to train a real-time detection model.

[0048] Step 3: Perform XRF spectrum element test on the suspected gangue to obtain element information, input the element information and the suspected gangue image into the fusion classification model, the fusion classification model then extracts element features through the element feature extraction module, extracts image features through the image feature extraction module, and performs matrix multiplication of image features and element features in the preliminary fusion module to form a similarity matrix, and then deeply fuses the similarity matrix with the image features to obtain the preliminary fusion features, and uses the preliminary fusion features to obtain the classification results of coal and gangue.

[0049] Specifically, the image feature extraction module is a convolutional neural network (CNN), which is a coal-rock fast extraction model composed of long- and short-range feature joint extraction blocks, such as Figure 5 As shown, the image feature extraction module extracts image features including:

[0050] (1) The attention matrix is ​​constructed using long- and short-distance features to capture the characteristics of coal and rock at different scales.

[0051] (2) The depth-wise separable convolutions of sizes 7×1 and 1×7 are used together to capture large-scale surface features. Moreover, the combination of 7×1 and 1×7 captures both horizontal and vertical features, which can avoid feature loss. Compared with directly using a 7×7 convolution kernel, the combination of 7×1 and 1×7 requires less computation and occupies less memory.

[0052] (3) Large-sized convolution kernels are prone to training difficulties. Therefore, when constructing the attention matrix, a 3×3 depthwise separable convolution is used to improve the model convergence speed and accuracy.

[0053] (4) Use point convolution to initially fuse features extracted from different distances, including:

[0054] f 1 = DWConv 7×1 ( x ) + DWConv 3×3 ( x ) + DWConv 1×7 ( x )

[0055] f 2 = BN(Conv 1×1 ( f 1))× x

[0056] f 3= f 2+ x

[0057] Among them, DWConv 7×1 Represents a depthwise separable convolution with a kernel size of 7×1, Conv 1×1 Represents a point convolution of size 1×1, DWConv 3×3 Depthwise Separable Convolution with a kernel size of 3×3, DWConv 1×7 Represents a depthwise separable convolution with a kernel size of 1×7; f 1 and f 2 is the intermediate variable value; f 3 is the feature output by the long-short distance feature joint extraction block;

[0058] Will f 3 is used as the input of the next long-short distance feature joint extraction block. After feature extraction by multiple long-short distance feature joint extraction blocks, the output image feature is finally obtained. .

[0059] Furthermore, the element feature extraction module is Fcs, which consists of multiple one-dimensional residual convolution modules, where 1DConv 3×1 represents a one-dimensional convolution kernel with a size of 3×1, BN represents normalization, and TanH represents a TanH activation function.

[0060] As an embodiment, the fusion classification model includes an element feature extraction module, an image feature extraction module, an initial fusion module and a deep fusion module. The image feature extraction module is a convolutional neural network, the element feature extraction module is Fcs, and the activation function in Fcs is a TanH function.

[0061] Furthermore, the training process of the fusion classification model is as follows: construct three datasets, dataset I contains only images, which is used to train the CNN module in the model; dataset II contains images and corresponding element data, which is used to train the entire model; dataset III contains images and corresponding element data, which is used to test CNN and the entire model.

[0062] Furthermore, the suspected gangue is subjected to XRF element testing to obtain element information, such as composition and corresponding content, such as iron, calcium, magnesium, silicon, phosphorus and other elements. The element information and the suspected gangue image are input into the fusion classification model, and then the element features are extracted by the element feature extraction module, and the image features are extracted by the image feature extraction module. The image features and element features are matrix multiplied in the initial fusion module to form a similarity matrix. The similarity matrix and the image features are deeply fused to obtain the initial fusion features, and the classification results of coal and gangue are obtained using the initial fusion features.

[0063] The fusion module structure is as follows Figure 2 As shown, the image features and element features are first multiplied to form a similarity matrix, including:

[0064] Similarity matrix calculation formula:

[0065]

[0066] in, are the features extracted from the image, are the features extracted from the elements.

[0067] Then, softmax is used to normalize the similarity matrix. Finally, the normalized similarity matrix is ​​multiplied by the image features to obtain the initial fusion features, where:

[0068] Softmax calculation formula:

[0069]

[0070] in, i, jare the row and column labels of the similarity matrix, e is a natural number, n is the number of columns of the similarity matrix; is the first i Line j The numeric value of the column.

[0071] Initial melting characteristic calculation formula:

[0072]

[0073] in, is the feature extracted from the image, and S is the similarity matrix.

[0074] The output layer of the fusion classification model uses the initial fusion features to output, obtains the classification results of coal and gangue, and outputs the probabilities of coal and gangue.

[0075] The coal and rock identification method based on the fusion of image and XRF spectrum information disclosed in the present invention is easy to cause feature loss and the related feature extraction module is difficult to train due to the different distribution characteristics of the two types of data. The similarity matrix proposed in the present invention can measure the similarity of the two types of features, which can help the module adjust its focus without causing feature loss. The method is simple to implement and has good results.

[0076] In addition, the long and short distance feature extraction blocks proposed in the present disclosure have no obvious shape features, only color and texture, because coal rock has significant self-similarity, that is, the local and the whole are very similar, and therefore coal rock has no obvious shape features, only color and texture, while conventional recognition tasks have color, texture and shape features. Therefore, compared with conventional recognition tasks, coal rock feature extraction is more difficult, and local feature extraction is more important. The long and short distance feature extraction blocks proposed in the present disclosure jointly use 1×7, 7×1, and 3×3 deep separation convolution kernels to extract features. Compared with traditional convolution kernels, the proposed module has stronger regional feature extraction capabilities and can effectively improve the classification effect.

[0077] Example 2

[0078] In one embodiment of the present disclosure, a coal rock identification system based on image and XRF spectrum information fusion is provided, comprising:

[0079] A data acquisition module, used to acquire the coal and rock images to be detected;

[0080] A preliminary detection module, used to input the coal-rock image into an image detection model to obtain suspicious coal gangue and its location information;

[0081] The deep recognition module is used to perform elemental testing on suspicious gangue to obtain elemental information, and input the elemental information and the image of suspicious gangue into the fusion classification model. The fusion classification model then extracts elemental features through the elemental feature extraction module and the image features through the image feature extraction module. The image features and elemental features are matrix multiplied in the initial fusion module to form a similarity matrix. The similarity matrix is ​​deeply fused with the image features to obtain the initial fusion features, and the classification results of coal and gangue are obtained using the initial fusion features.

[0082] Among them, the data acquisition module includes an image acquisition module and a grabbing module. The image acquisition module includes a camera installed directly above the conveyor belt. The camera collects images of coal and coal gangue on the conveyor belt in real time and transmits the images to the central processing module. The central processing module then transmits the data to be detected to the intelligent detection module.

[0083] An image detection model and a fusion classification model are deployed in the intelligent detection module. The training of the gangue image detection model includes the following steps: ① collecting gangue images on the conveyor belt and annotating the gangue and coal in the images; ② using the annotated data to create a training set, a validation set and a test set; ③ using the training set, the validation set and the test set to train a real-time detection model.

[0084] The image detection model takes the image as input and can provide feedback to the central processing module on whether there is gangue in the image and its location information.

[0085] The fusion classification model takes image data and element data as input and can feed back the gangue probability information to the central processing module.

[0086] The grabbing module receives the grabbing instruction and the position information of the suspected gangue object from the central processing module, and grabs the suspected gangue object from the conveyor belt to the element to be tested area.

[0087] Furthermore, the functions of the central processing module include: ① controlling the image acquisition module to acquire images, ② controlling the element test module to acquire data, ③ controlling the grasping module to grasp objects, and ④ sending the data to be detected to the intelligent detection module to obtain the detection results.

[0088] Furthermore, the data acquisition module also includes an element testing module, which includes an X-ray fluorescence spectrometer (XRF) that can test the element information of the object being tested. After completing the test, the element testing module will send the element information to the central processing module.

[0089] Furthermore, the implementation process of the coal rock identification system based on the fusion of image and XRF spectrum information includes:

[0090] (1) The central processing module sends image acquisition instructions to the image acquisition module;

[0091] (2) The image acquisition module acquires the conveyor belt image and sends it to the central processing module;

[0092] (3) The central processing module sends the image to the intelligent detection module;

[0093] (4) The intelligent detection module calls the image detection model to detect objects in the image, determine whether there is suspicious gangue and its location information, and transmit the result to the central processing module. If there is suspicious gangue, the result also includes the gangue image;

[0094] (5) The central processing module judges the detection results. If there is no suspicious gangue module, the process ends. If there is suspicious gangue, the central processing module sends the suspicious gangue location information to the grabbing module.

[0095] (6) The grabbing module grabs the suspected gangue on the conveyor belt to the element testing area according to the gangue location information, and transmits the grabbing result to the central processing module;

[0096] (7) The central processing module sends element testing instructions to the element testing module.

[0097] (8) The XRF spectrum element test module performs element composition test on the suspected coal gangue and transmits the results to the central processing module;

[0098] (9) The central processing module sends the element test results and coal gangue images to the intelligent detection module;

[0099] (10) The intelligent detection module mobilizes the fusion classification model of coal gangue to carry out coal gangue classification and transmits the classification results to the central processing module;

[0100] (11) If the classification result is coal gangue, the central processing module mobilizes the grabbing module to put the coal gangue into the coal gangue area, otherwise the central processing module mobilizes the grabbing module to send the coal back to the conveyor belt.

[0101] As an embodiment, the method executed by the system of the present disclosure can also be applied to the accurate classification of coal gangue and coal on site underground.

[0102] The system includes a central processing module, an image acquisition module, an intelligent detection module and an element testing module.

[0103] (1) The central processing module controls the other three modules;

[0104] (2) The image acquisition module is used to collect images of the test object on site;

[0105] (3) The intelligent detection module is used for intelligent classification, including an image-based image detection model and a fusion classification model; the image detection model can be a conventional convolutional neural network, such as the common ResNet-18; the image-based detection model training includes the following steps: ① Collect images of coal gangue on the conveyor belt and label the coal gangue and coal in the image; ② Use the labeled data to create a training set, a validation set, and a test set; ③ Use the training set, the validation set, and the test set to train a real-time detection model.

[0106] (4) The element test module is used to test the elemental composition of the object being tested.

[0107] The real-time process steps implemented by the central processing module, image acquisition module, intelligent detection module and element testing module include:

[0108] (1) The central processing module mobilizes the image acquisition module to collect images and pass the results to the intelligent detection model;

[0109] (2) The intelligent detection module mobilizes the image-based detection model to classify the image and passes the results to the central processing module;

[0110] (3) The central processing module judges the recognition result. If the recognition probability is high, such as greater than 80%, it indicates that the image-based detection model is relatively accurate, and the detection result is given directly. Otherwise, it indicates that the image-based detection model is not confident in the recognition result, and the central processing module will mobilize the element testing module to perform elemental composition testing on the object to be tested.

[0111] (4) After the element testing module completes the test, it sends the element composition information to the central processing module;

[0112] (5) The central processing module sends the element information and image to the intelligent detection model;

[0113] (6) The intelligent detection model uses element information and images as input to mobilize the fusion classification model for classification, and finally sends the classification results to the central processing module;

[0114] (7) The central processing module displays the final classification results.

[0115] Example 3

[0116] In one embodiment of the present disclosure, a computer program product is provided, including a computer program, which, when executed by a processor, implements the coal and rock identification method based on the fusion of image and XRF spectrum information.

[0117] Example 4

[0118] In one embodiment of the present disclosure, a non-transitory computer-readable storage medium is provided, wherein the non-transitory computer-readable storage medium is used to store computer instructions. When the computer instructions are executed by a processor, the coal and rock identification method based on the fusion of image and XRF spectral information is implemented.

[0119] Example 5

[0120] In one embodiment of the present disclosure, an electronic device is provided, including: a processor, a memory, and a computer program; wherein the processor is connected to the memory, and the computer program is stored in the memory. When the electronic device is running, the processor executes the computer program stored in the memory, so that the electronic device executes the coal rock identification method based on the fusion of image and XRF spectral information.

[0121] The present disclosure is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present disclosure. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of the processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0122] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0123] Although the above describes the specific implementation methods of the present disclosure in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present disclosure. Technical personnel in the relevant field should understand that on the basis of the technical solution of the present disclosure, various modifications or variations that can be made by those skilled in the art without creative work are still within the scope of protection of the present disclosure.

Claims

1. A coal-rock identification method based on the fusion of image and XRF spectrum information, characterized in that: include: Acquire coal and rock images to be detected; Inputting the coal-rock image into an image detection model to obtain suspicious coal gangue and its location information; Conduct XRF spectrum element test on suspicious gangue to obtain element information, input the element information and the image of suspicious gangue into the fusion classification model, the fusion classification model extracts element features through the element feature extraction module, extracts image features through the image feature extraction module, performs matrix multiplication of image features and element features in the initial fusion module to form a similarity matrix, deeply fuses the similarity matrix with the image features to obtain the initial fusion features, and uses the initial fusion features to obtain the classification results of coal and gangue; The image features and element features are matrix-multiplied in the initial fusion module to form a similarity matrix, and the similarity matrix is ​​fused with the image features to obtain the initial fusion features, including: the image features and element features are first multiplied to form a similarity matrix, and then the similarity matrix is ​​normalized using softmax. Finally, the normalized similarity matrix is ​​multiplied with the image features to obtain the initial fusion features; Similarity matrix calculation formula: Softmax calculation formula: in, i, j are the row and column labels of the similarity matrix, e is a natural number, n is the number of columns of the similarity matrix, is the first i Line j The value of the column; are the features extracted from the image, are the features extracted from the elements.

2. The coal-rock identification method based on image and XRF spectrum information fusion according to claim 1, characterized in that: The image feature extraction module is a convolutional neural network, which includes: using point convolution to initially fuse features extracted from different distances: f 1 = DWConv 7×1 ( x ) + DWConv 3×3 ( x ) + DWConv 1×7 ( x ) f 2 = BN(Conv 1×1 ( f 1))× x f 3 = f 2 + x Among them, DWConv 7×1 Represents a depthwise separable convolution with a kernel size of 7×1, Conv 1×1 Represents a point convolution of size 1×1, DWConv 3×3 ( x ) represents a depth-separated convolution with a kernel size of 3×3, DWConv 1×7 ( x ) represents a depthwise separable convolution with a kernel size of 1×7; f 1 and f 2 is the intermediate variable value; f 3 is the feature output by the long-short distance feature joint extraction block; The element feature extraction module is Fcs, and the activation function in Fcs is the TanH function.

3. The coal-rock identification method based on image and XRF spectrum information fusion according to claim 1, characterized in that: The similarity matrix and image features are deeply fused to obtain the initial fusion features, including: in, is the feature extracted from the image, and S is the similarity matrix.

4. The coal-rock identification method based on image and XRF spectrum information fusion according to claim 1, characterized in that: Collect coal and rock images on the conveyor belt, input the acquired coal and rock images into the image detection model, obtain the image and position information of suspicious coal gangue, use the grabbing module to grab the suspicious coal gangue to the element testing area according to the position information, and perform element testing.

5. A coal-rock identification system based on the fusion of image and XRF spectrum information, which executes the coal-rock identification method based on the fusion of image and XRF spectrum information as claimed in any one of claims 1 to 4, characterized in that: include: A data acquisition module, used to acquire the coal and rock images to be detected; A preliminary detection module, used to input the coal-rock image into an image detection model to obtain suspicious coal gangue and its location information; The deep recognition module is used to perform elemental testing on suspicious gangue to obtain elemental information, and input the elemental information and the suspicious gangue image into the fusion classification model. The fusion classification model extracts elemental features through the elemental feature extraction module and image features through the image feature extraction module, and performs matrix multiplication of the image features and the elemental features in the initial fusion module to form a similarity matrix. The similarity matrix and the image features are deeply fused to obtain the initial fusion features, and the classification results of coal and gangue are obtained using the initial fusion features.

6. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the method for identifying coal and rock based on the fusion of image and XRF spectrum information as described in any one of claims 1 to 4 is implemented.

7. A non-transitory computer-readable storage medium, characterized in that: The non-transitory computer-readable storage medium is used to store computer instructions. When the computer instructions are executed by the processor, the coal and rock identification method based on the fusion of image and XRF spectrum information as described in any one of claims 1 to 4 is implemented.

8. An electronic device, characterized in that: include: A processor, a memory and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory so that the electronic device executes the coal rock identification method based on the fusion of image and XRF spectrum information as described in any one of claims 1-4.

Images