Coal gangue grading and sorting method and system based on LIBS and image analysis

Abstract

The invention discloses a coal gangue grading and sorting method and system based on LIBS (laser-induced breakdown spectroscopy) and image analysis, and relates to the technical field of coal gangue sorting, the method comprises the following steps: acquiring spectral data of each sample in a mixed coal flow comprising coal briquettes and coal gangue by using an LIBS technology; comparing the obtained spectral data with a pre-established coal briquette spectral database, preliminarily identifying coal briquettes and coal gangue according to a comparison result, and completing preliminary sorting; obtaining a visible light image of the separated mixed coal flow, preprocessing the image, inputting the preprocessed image into a coal gangue detection model based on an improved Mask RCNN, identifying and segmenting all coal gangues in the image, and obtaining the equivalent particle size of each coal gangue; and according to a preset particle size range, the coal gangue is divided into different grades, and the coal gangue in the mixed coal flow is classified and sorted. According to the invention, accurate identification of the coal gangue and grading sorting of the coal gangue can be realized.

Description

Technical Field

[0001] The present invention relates to the technical field of coal gangue sorting, and in particular to a coal gangue grading and sorting method and system based on LIBS and image analysis. Background Art

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

[0003] As one of the main energy sources at present, coal plays an irreplaceable role. Coal gangue is mining waste in coal mining, accounting for up to 15% of coal production. Since coal briquettes and coal gangue have significant differences in material composition, application fields, appearance characteristics and environmental impact, coal gangue is relatively harder and has a relatively more complex composition. It has a lower carbon content, contains more impurities and lower calorific value, and is usually not used directly as fuel. Furthermore, a high proportion of coal gangue mixed into the coal flow will lead to a reduction in the quality of raw coal and commercial coal. Therefore, coal gangue sorting, as a coal pre-combustion treatment technology, is an indispensable link in coal mine production and one of the effective methods to improve coal quality.

[0004] Traditionally, gangue sorting mainly relies on manual labor, which is labor-intensive, has a poor working environment, and has the risk of misselection or omission. Gangue may contain flammable, explosive, or toxic and hazardous substances, and manual sorting has safety risks and may cause environmental pollution. With the increase in coal mining and the improvement of environmental protection requirements, the amount of gangue processed is also increasing. Manual sorting is difficult to adapt to the needs of large-scale processing and cannot meet the requirements of modern industrial production for efficiency and quality.

[0005] With the development of science and technology, coal gangue sorting technology is becoming more and more intelligent and automated. By introducing intelligent sorting technology and strengthening mechanization development, it is now possible to achieve intelligent identification and screening of coal gangue by combining advanced technologies such as image processing, machine learning and automatic control. Through automated mechanical sorting and control, the sorting efficiency can be improved to a certain extent, thereby reducing labor costs, reducing human errors, avoiding possible work-related accidents during manual sorting, and ensuring the safety of workers. However, the existing intelligent coal gangue sorting methods still have certain problems, as follows: (1) Existing methods usually only use a single intelligent technology to sort coal gangue. Due to the complex composition and properties of coal gangue, a single identification method is often difficult to adapt to the needs of different situations and it is difficult to ensure the accuracy of identification and sorting, which often causes some coal gangue to be missed and mixed into the coal flow; (2) Considering that the resource utilization methods of coal gangue are gradually becoming diversified and harmless, coal gangue can be used in many fields after proper treatment. However, most of the existing methods only sort out the coal gangue without grading it for recycling and reuse, resulting in a certain amount of resource waste. Summary of the invention

[0006] In order to solve the deficiencies of the above-mentioned prior art, the present invention provides a coal gangue grading and sorting method and system based on LIBS and image analysis, which combines LIBS (Laser-induced breakdown spectroscopy) technology and image processing technology for analysis, that is, firstly using LIBS technology to distinguish coal lumps and coal gangue, and separate the coal lumps from the mixed coal flow, and then using image processing technology to re-sort and classify the coal gangue by particle size, and classify the coal gangue according to the particle size, so as to meet the needs of intelligent and accurate sorting of coal gangue and graded utilization of coal gangue, thereby realizing the recycling and reuse of coal gangue and reducing resource waste.

[0007] In a first aspect, the present invention provides a method for grading and sorting coal gangue based on LIBS and image analysis.

[0008] A method for grading and sorting coal gangue based on LIBS and image analysis, comprising: Using LIBS technology, the spectral data of each sample in the mixed coal flow including coal lumps and coal gangue are obtained; The acquired spectral data is compared with the pre-established coal block spectral database, and the coal blocks and coal gangue are preliminarily identified based on the comparison results; Screen and separate the initially identified coal blocks into mixed coal streams, and complete the initial sorting of coal gangue in the mixed coal streams; Obtain a visible light image of the mixed coal flow after separation, perform preprocessing on the image, input the preprocessed image into the coal gangue detection model based on the improved Mask RCNN, identify and segment all the coal gangue in the image, and obtain the equivalent particle size of each coal gangue; According to the preset particle size range, the gangue is divided into different levels, and the gangue in the mixed coal flow is graded and sorted.

[0009] In a second aspect, the present invention provides a coal gangue grading and sorting system based on LIBS and image analysis.

[0010] A coal gangue grading and sorting system based on LIBS and image analysis, comprising: A spectral data acquisition module, used to acquire spectral data of each sample in a mixed coal flow including coal lumps and coal gangue using LIBS technology; A coal block and coal gangue preliminary identification module is used to compare the acquired spectral data with the pre-established coal block spectral database, and preliminarily identify the coal block and coal gangue according to the comparison results; The gangue preliminary sorting module is used to screen the preliminarily identified coal blocks and separate the mixed coal flow, completing the preliminary sorting of the gangue in the mixed coal flow; The gangue detection module is used to obtain the visible light image of the mixed coal flow after separation, and preprocess the image. The preprocessed image is input into the gangue detection model based on the improved Mask RCNN to identify and segment all the gangue in the image and obtain the equivalent particle size of each gangue. The gangue grading and sorting module is used to divide the gangue into different levels according to the preset particle size range and to grade and sort the gangue in the mixed coal flow.

[0011] In a third aspect, the present invention further provides an electronic device comprising a memory and a processor, and computer instructions stored in the memory and executed on the processor, wherein when the computer instructions are executed by the processor, the steps of the method described in the first aspect are completed.

[0012] In a fourth aspect, the present invention further provides a computer-readable storage medium for storing computer instructions, which, when executed by a processor, complete the steps of the method described in the first aspect.

[0013] One or more of the above technical solutions have the following beneficial effects: 1. The present invention provides a method and system for grading and sorting coal gangue based on LIBS and image analysis. By combining multiple recognition technologies: LIBS technology and image processing technology, the coal gangue in the mixed coal flow is identified, and the coal gangue is graded and sorted according to the recognition results, so as to ensure that the recognition and sorting process is more accurate, reduce misjudgment and missed judgment, and avoid the situation where some coal gangue is missed due to the use of only a single intelligent technology to sort coal gangue.

[0014] 2. The gangue grading and sorting method proposed in the present invention first uses LIBS technology to perform preliminary identification of coal blocks and gangue, so as to distinguish coal blocks from gangue, and separate the coal blocks from the mixed coal flow according to the preliminary identification results; considering that only using a single LIBS technology to distinguish coal blocks and gangue may result in identification errors or omissions, the present invention also proposes a secondary identification method based on an image recognition algorithm, which uses an improved network model to re-sort and classify the particle size of the gangue after preliminary sorting, so as to meet the requirements of gangue. The demand for intelligent and precise sorting is met, and by using the method of spectral recognition followed by image recognition, it is possible to ensure the detection and recognition accuracy while avoiding the need for large amounts of data processing for image recognition, thereby further reducing the computational cost of image recognition and improving the recognition and sorting efficiency; in addition, considering that coal gangue of different particle sizes can be applied to different fields, the present invention calculates the particle size of the identified coal gangue and classes the coal gangue according to the particle size, thereby meeting the demand for graded utilization of coal gangue, and further realizing the recycling and reuse of coal gangue and reducing resource waste.

[0015] 3. The automated and intelligent sorting method proposed in the present invention greatly improves the sorting efficiency, reduces manual intervention, and reduces labor intensity; the graded recovery of coal gangue through image technology realizes the maximum utilization of resources and helps promote the construction of green mines. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0017] Figure 1 This is a flow chart of a method for grading and sorting coal gangue based on LIBS and image analysis according to an embodiment of the present invention; Figure 2 This is a flow chart for preliminary identification of coal lumps and coal gangue in an embodiment of the present invention; Figure 3 Schematic diagram of the structure of a gangue detection model based on improved Mask RCNN in an embodiment of the present invention. DETAILED DESCRIPTION

[0018] It should be noted that the following detailed descriptions are exemplary only, are intended to describe specific embodiments, are intended to provide further explanation of the present invention, and are not intended to limit exemplary embodiments according to the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those of ordinary skill in the art to which the present invention belongs. In addition, it should also 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.

[0019] Embodiment 1 This embodiment provides a coal gangue classification and sorting method based on LIBS and image analysis, such as Figure 1 As shown, the specific steps include: Step S1, using LIBS technology to obtain the spectral data of each sample in the mixed coal flow including coal blocks and coal gangue. Specifically, LIBS technology can form plasma by focusing ultrashort pulse laser on the sample surface, and the material composition and content of the sample can be determined by acquiring and analyzing the plasma emission spectrum. In this embodiment, after the coal mining is completed, the mixed coal flow of mined coal blocks and coal gangue is sent to the sorting system through a conveyor belt and other devices, and the mixed coal flow of coal blocks and coal gangue enters the sorting process. The spectral data of each substance (or sample) in the mixed coal flow is obtained by using LIBS technology, that is, the LIBS laser is started, and the coal blocks and coal gangue in the mixed coal flow are laser-induced to break down, and then the plasma emission spectrum generated by the laser induction is collected by the multi-spectral spectrometer to form the spectral data of the detected sample (including coal blocks and coal gangue), and the spectral data is the spectral intensity.

[0020] Furthermore, the spectral data obtained by the above method is multi-channel spectral data. Since each channel in the multi-channel spectrometer is independent, specifically, the wavelength range and spectral resolution of each channel are different, it is necessary to perform independent normalization processing on each channel, that is, after obtaining the multi-channel spectral data of each sample in the mixed coal flow, the spectral data of each channel is independently normalized. Specifically, the normalization process of each channel can be obtained by the following calculation: ; in, is the normalized spectral intensity in each channel, is the spectral intensity value of each group of data after the background spectrum is removed, is the minimum spectral intensity in each set of data, is the maximum spectral intensity in each set of data.

[0021] Step S2: compare the acquired spectral data with the pre-established coal block spectral database, and preliminarily identify the coal blocks and coal gangue according to the comparison results.

[0022] Specifically, the above-mentioned pre-established coal block spectrum database is constructed through preliminary investigation and analysis of coal mining. Through the preliminary investigation, coal block samples (i.e., coal samples) in the coal mining area are obtained, and the characteristics of the coal blocks in the coal mining area are fully investigated, so as to lay the foundation for the subsequent creation of the coal block spectrum database. Specifically, during the coal mining process, coal samples are collected from the coal mining area in the coal mine, and the surface of the collected coal block samples is excited using LIBS technology, and the multi-channel spectrum data corresponding to each coal block sample is collected and obtained. Then, for the multi-channel spectrum data of each sample, the spectrum data of each channel is independently normalized, and finally, the normalized spectrum data is analyzed and processed, and based on the spectrum data of each channel of the processed coal block (such as spectrum intensity, etc.), a coal block spectrum database (also referred to as coal spectrum database) is formed.

[0023] like Figure 2 As shown, after the coal mining is completed, the spectral data obtained in step S1 is compared with the above-mentioned pre-established coal block spectral database. The comparison method is: the obtained sample spectral data is compared with the spectral data in the coal block spectral database one by one by channel, and the difference between the sample spectral data and the coal block spectral data of each channel is calculated; wherein, if the difference in the spectral data of each channel of the sample is not less than the set threshold corresponding to the current channel, the sample is determined to be a coal block; conversely, if the difference in the spectral intensity of a certain channel in the sample is less than the set threshold corresponding to the current channel, the sample is determined to be coal gangue. Furthermore, accurate identification is performed based on the above-mentioned spectral database, and identification and determination as coal block or coal gangue are performed based on the comparison results. The process can be expressed as: First, set up a pre-established coal spectral database D , which is actually a two-dimensional array, D [ k, j ] indicates the first k The first j The spectral intensity of each channel; at the same time, set the spectral data of the sample S , the spectral data S is a two-dimensional array, S [ i, j ] indicates the i The sample in j Spectral intensity of each channel; preset threshold value corresponding to each channel , in order to determine whether the spectral intensity difference is within the error range.

[0024] Secondly, for each sample i and each coal block in the coal block spectrum database k , calculate the spectral intensity difference in each channel, which can be expressed as: ; in, S [ i, j ] indicates the i Sample No. j The spectral intensity of the channel, D [ k, j ] indicates the k Coal block j The spectral intensity of each channel, K represents the number of coal blocks in the coal block spectrum database.

[0025] Finally, for each sample i , if the difference of the spectral data of each channel ΔS [ i, j ] are not less than the set threshold corresponding to the current channel , it can be determined to be a coal block; otherwise, if there is a difference in the spectral intensity of a certain channel ΔS [ i, j ] is less than the set threshold corresponding to the current channel , it is determined to be coal gangue.

[0026] Step S3, screen the initially identified coal lumps and separate the mixed coal flow, completing the initial sorting of the coal gangue in the mixed coal flow. Specifically, according to the determination result of the above step S2, the coal lumps are separated from the mixed coal flow, and the remaining mixed coal flow including coal gangue is transmitted to the next classification process.

[0027] In the above method, firstly, coal blocks and coal gangue in the coal mining area are collected. Preferably, the samples can be pre-processed by crushing, screening, etc. as needed to better expose their chemical composition, thereby improving the resolution efficiency; secondly, the surface of the collected coal block and coal gangue samples is excited by LIBS technology, that is, the laser source is controlled to emit high-energy pulsed laser, irradiated to the sample surface, and the laser interacts with the sample surface to generate plasma and emit spectral signals; then, after collecting the spectral data, the above data normalization processing and other analytical processing are performed on it, that is, the spectral data emitted by the plasma is collected by a spectrometer, and it is converted into an electrical signal, and the spectral data is processed and analyzed using an existing algorithm to identify the different elemental compositions and contents of the coal blocks and coal gangue; finally, according to the spectral data obtained by LIBS, the spectral database and pattern recognition algorithm are used to analyze the LIBS spectral data, identify and distinguish the coal blocks and coal gangue. In the above method, the coal blocks can be separated from the mixed coal flow according to the preliminary identification results. Furthermore, considering that only one method may lead to false detection and missed detection when identifying gangue, and the above method alone cannot accurately classify the particle size of gangue, this embodiment also provides a secondary recognition method based on an image recognition algorithm, including: Step S4, obtain a visible light image of the mixed coal flow after separation, and perform preprocessing operations on the image, input the preprocessed image into the coal gangue detection model based on the improved Mask RCNN, identify and segment all the coal gangue in the image, and obtain the equivalent particle size of each coal gangue.

[0028] Specifically, the remaining mixed coal flow (mainly including gangue) after LIBS screening is imaged, and the corresponding image data is obtained using a high-definition camera. Secondly, the collected images are preprocessed by cropping, denoising, graying, etc. to improve image quality and reduce interference factors. Then, the image is processed by a deep learning algorithm, that is, the preprocessed image is input into the gangue detection model based on the improved Mask RCNN neural network to identify and segment all the gangue in the image, and obtain the equivalent particle size of each gangue.

[0029] In this embodiment, an improved Mask RCNN network is used as a gangue detection model, and the gangue detection model is used to detect and identify gangue in the input image, and each gangue area is segmented, and the equivalent particle size of the gangue is calculated based on the segmented gangue area. Figure 3 As shown in the figure, the Mask RCNN network model includes a feature extraction module, a region generation module, a ROI Align module, and a parallel Class&Box branch and Mask branch, which are connected in sequence.

[0030] Among them, the feature extraction module includes a ResNet residual block, a dual-channel SE module and a self-attention module: the ResNet residual block is used to preliminarily extract the feature map; the dual-channel SE module is used to extract the channel weights of the input feature map, and output the feature map weighted based on the channel weights; the self-attention module is used to extract the self-attention weights of the input feature map, and output the feature map weighted based on the self-attention weights; the feature map after channel and space weighting by the dual-channel SE module and the self-attention module are added and fused with the feature map output by the ResNet residual block, and the final extracted feature map is output.

[0031] Specifically, for the feature extraction module, the ResNet residual network (i.e., ResNet residual block) is first used to preliminarily extract the features of the input image, i.e., extract the feature map. The ResNet residual block adopts a residual block-based structure, including two convolutional layers, a batch normalization layer, and a ReLU activation function, wherein the convolutional layer is used to extract features, the batch normalization layer is used to stabilize the training process, and the ReLU activation function is used to increase nonlinearity.

[0032] Secondly, in order to further improve the accuracy of detection and recognition, this embodiment adds a dual-channel SE module (SE, Squeeze-and-Excitation, global information embedding and adaptive recalibration) on the basis of the ResNet residual block to form a dual-channel SE residual block, which combines the residual learning of ResNet and the feature recalibration capability of the SE module. Among them, the feature map output by the ResNet residual block is input into the dual-channel SE module. In the dual-channel SE module, the feature map is first compressed into a dual-channel channel descriptor by global average pooling, and then the dependency between channels is learned through a fully connected layer. Then, the weight of each channel is obtained by the Sigmoid activation function, that is, the channel weight is obtained, and finally, the channel weight is multiplied by the feature map of the input dual-channel SE module to realize the recalibration of the channel features and output the weighted feature map. The above-mentioned dual-channel SE module is used to perform channel-level recalibration, generate channel weights, and obtain a feature map weighted by the channel weights.

[0033] At the same time, on the basis of the dual-channel SE module, a self-attention module is set up to introduce a self-attention mechanism, through which the feature map is weighted at the spatial level. The feature map output by the ResNet residual block is input to the self-attention module, and the self-attention weight of the input feature map is learned and extracted through the self-attention mechanism. The self-attention weight is multiplied with the feature map input to the self-attention module to achieve spatial recalibration and output the feature map after self-attention weighting.

[0034] Finally, in order to maintain the properties of the residual block, the feature map output by the ResNet residual block is added and fused with the feature map after channel and spatial weighting by the dual-channel SE module and the self-attention module respectively to form a residual connection and output the final extracted feature map.

[0035] By combining the above-mentioned ResNet residual block, dual-channel SE module and self-attention module, the comprehensiveness and effectiveness of feature extraction are guaranteed, and the accuracy of subsequent coal gangue segmentation is further improved.

[0036] Secondly, the extracted feature map is input into the region generation module, which uses the RPN network (Region Proposal Network) to output a series of candidate boxes. The output candidate boxes are combined with the feature map, and the part of the feature map corresponding to the candidate box is cropped and input into the RoIAlign layer to generate a candidate box feature map with the same size.

[0037] Through the above steps, the image is divided into multiple gangue feature sub-regions (i.e., feature regions / feature maps corresponding to candidate boxes). The RoIAlign layer receives the feature maps corresponding to these candidate boxes, maps the candidate box coordinates, obtains the corresponding feature map coordinates, and calculates the feature value of each sampling point / pixel point based on the four closest pixels using bilinear interpolation, and then performs average weighting to make the extracted features more obvious.

[0038] Finally, the feature map output by the ROI Align layer is input into the parallel Class&Box branch and Mask branch respectively, wherein the Class&Box branch is used to obtain the position and category information of the target object (i.e., gangue) in the image, and the Mask branch is used to obtain the segmentation information of the target object (i.e., gangue) in the image, and finally outputs the position and category of the gangue in the identified image, and accurately segments the identified gangue area. Among them, this embodiment adds a Mask branch to predict the mask of the object, thereby realizing instance segmentation. Specifically, the Mask branch includes a plurality of convolutional layers with different convolutional kernel sizes, which are used to extract and process the input feature map, and further extract features using the convolutional layer, and finally outputs a two-dimensional mask corresponding to the size of the RoI (Region of Interest, region of interest), wherein each pixel in the mask indicates whether the point belongs to the extracted gangue area, and 0 and 1 are usually used to represent the non-target area and the target area, thereby outputting the accurately segmented gangue area.

[0039] Furthermore, the loss function of the above Mask RCNN network model consists of four parts: classification error loss, detection error loss, IOU error loss and improved segmentation error loss. Among them, the classification error loss and detection error loss are both calculated using the cross entropy loss function; the IoU error loss is used to evaluate the degree of overlap between the detection frame and the real object, so as to measure the quality of the generated detection frame. By increasing the IoU error loss, the network can pay more attention to the overlap between the predicted results and the real annotations, further ensuring the detection accuracy of the detection frame, and laying the foundation for the subsequent detection of coal gangue particle size classification in the frame. The IoU error loss can be expressed as: ; in, B and denote the true bounding box and the predicted bounding box respectively, Intersection represents the area of ​​the intersection region, Union Represents the area of ​​the union region.

[0040] In addition, the segmentation error loss is the loss of the Mask branch, which is calculated using the average binary cross entropy loss function. This loss function only calculates the mask loss corresponding to its true category, while the mask loss of other categories does not contribute to the total loss, thereby avoiding inter-class competition problems and allowing each category to generate a mask independently. The segmentation error loss function is: .

[0041] In order to further improve the accuracy of recognition and segmentation, the segmentation error loss is further improved. The boundary weighted loss function is introduced into the traditional segmentation error loss function. The boundary information is weighted to improve the accuracy of coal gangue image segmentation. The improved segmentation error loss is: ; in, represents the average binary cross entropy loss function, represents the boundary weighted loss function, is the weight coefficient, which is used to adjust the boundary of the segmentation result B The extent of its impact on the overall loss; R For all segmented areas; Indicates the distance from each pixel to the boundary; y is the ground-truth (i.e., the real label) after binarization; is the predicted segmentation result after binarization (i.e., predicted label); p is the target location.

[0042] Furthermore, all the gangue in the image is identified and segmented based on the above network model, and for each segmented gangue area, the equivalent particle size of each gangue is calculated. In this embodiment, in the process of analyzing and detecting the particle size of the gangue, the equivalent particle size of the gangue is used as the classification standard for the particle size classification of the gangue, and the perimeter of the maximum projection surface of the segmented gangue area is obtained. P and area A , to calculate the major axis of the ellipse circumscribing the region a and short axis b The equivalent particle size of coal gangue can be calculated by using these two parameters. d The calculation formula is as follows: ; ; ; in, P is the perimeter of the maximum projection surface of coal gangue; A It is the area of ​​the maximum projection surface of coal gangue.

[0043] Step S5: classify the gangue into different levels according to the preset particle size range, and classify and sort the gangue in the mixed coal flow.

[0044] Specifically, according to the recycling and utilization of coal gangue, the standard of coal gangue particle size classification is defined, and different grades correspond to different particle size ranges. Based on the identified and calculated equivalent particle size of coal gangue d , the gangue is divided into different levels, and on this basis, based on the equivalent particle size of the gangue, the gangue is graded and sorted by corresponding devices, and the graded gangue is transported to different areas for subsequent recycling to avoid waste of resources.

[0045] Embodiment 2 This embodiment provides a coal gangue grading and sorting system based on LIBS and image analysis, including: A spectral data acquisition module, used to acquire spectral data of each sample in a mixed coal flow including coal lumps and coal gangue using LIBS technology; A coal block and coal gangue preliminary identification module is used to compare the acquired spectral data with the pre-established coal block spectral database, and preliminarily identify the coal block and coal gangue according to the comparison results; The gangue preliminary sorting module is used to screen the preliminarily identified coal blocks and separate the mixed coal flow, completing the preliminary sorting of the gangue in the mixed coal flow; The gangue detection module is used to obtain the visible light image of the mixed coal flow after separation, and preprocess the image. The preprocessed image is input into the gangue detection model based on the improved Mask RCNN to identify and segment all the gangue in the image and obtain the equivalent particle size of each gangue. The gangue grading and sorting module is used to divide the gangue into different levels according to the preset particle size range and to grade and sort the gangue in the mixed coal flow.

[0046] Embodiment 3 This embodiment provides an electronic device, including a memory and a processor, and computer instructions stored in the memory and running on the processor. When the computer instructions are run by the processor, the steps in the coal gangue grading and sorting method based on LIBS and image analysis as described above are completed.

[0047] Embodiment 4 This embodiment also provides a computer-readable storage medium for storing computer instructions. When the computer instructions are executed by a processor, the steps in the coal gangue grading and sorting method based on LIBS and image analysis as described above are completed.

[0048] The steps involved in the above embodiments 2 to 4 correspond to the method embodiment 1, and the specific implementation methods can refer to the relevant description part of embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood to include any medium that can store, encode or carry an instruction set for execution by a processor and enable the processor to execute any method in the present invention.

[0049] Those skilled in the art should understand that the modules or steps of the present invention described above can be implemented by a general-purpose computer device, or alternatively, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. The present invention is not limited to any specific combination of hardware and software.

[0050] The above description is only a preferred embodiment of the present invention. Although the specific implementation mode of the present invention is described in conjunction with the accompanying drawings, it is not a limitation of the protection scope of the present invention. Those skilled in the art should understand that on the basis of the technical solution of the present invention, various modifications or variations that can be made by those skilled in the art without creative work are still within the protection scope of the present invention.

Claims

1. A coal gangue grading and sorting method based on LIBS and image analysis, characterized in that: include: Using LIBS technology, the spectral data of each sample in the mixed coal flow including coal lumps and coal gangue are obtained; The acquired spectral data is compared with the pre-established coal block spectral database, and the coal blocks and coal gangue are preliminarily identified based on the comparison results; Screen and separate the initially identified coal blocks into mixed coal streams, and complete the initial sorting of coal gangue in the mixed coal streams; Obtain a visible light image of the mixed coal flow after separation, perform preprocessing on the image, input the preprocessed image into the coal gangue detection model based on the improved Mask RCNN, identify and segment all the coal gangue in the image, and obtain the equivalent particle size of each coal gangue; According to the preset particle size range, the gangue is divided into different levels, and the gangue in the mixed coal flow is graded and sorted.

2. A method for grading and sorting coal gangue based on LIBS and image analysis as claimed in claim 1, characterized in that: The construction process of the coal block spectrum database is as follows: During the coal mining process, several coal block samples are collected in the coal mine, and all coal block samples are scanned using LIBS technology to obtain multi-channel spectral data corresponding to each coal block sample; For the multi-channel spectral data of each coal block sample, the spectral data of each channel is independently normalized; Based on the normalized multi-channel spectral data, a coal block spectral database is constructed.

3. A method for grading and sorting coal gangue based on LIBS and image analysis as claimed in claim 1, characterized in that: The acquired spectral data is compared with the pre-established coal block spectral database, including: The acquired sample spectrum data is compared with the spectrum data in the coal block spectrum database one by one according to the channels, and the difference between the sample spectrum data and the coal block spectrum data in each channel is calculated; If the difference of the spectral data of each channel of the sample is not less than the set threshold corresponding to the current channel, the sample is determined to be a coal block; On the contrary, if the difference of the spectral intensity of a certain channel in the sample is less than the set threshold corresponding to the current channel, the sample is determined to be coal gangue.

4. A method for grading and sorting coal gangue based on LIBS and image analysis as claimed in claim 1, characterized in that: The gangue detection model adopts an improved Mask RCNN, and the improved Mask RCNN network model includes a feature extraction module, a region generation module, a ROI Align module, and a parallel Class&Box branch and a Mask branch connected in sequence; The feature extraction module includes a ResNet residual block, a dual-channel SE module and a self-attention module, wherein the ResNet residual block is used to preliminarily extract a feature map; the dual-channel SE module is used to extract the channel weight of the input feature map and output a feature map weighted by the channel weight; the self-attention module is used to extract the self-attention weight of the input feature map and output a feature map weighted by the self-attention weight; The feature maps after channel and spatial weighting are respectively processed by the dual-channel SE module and the self-attention module, and then added and fused with the feature maps output by the ResNet residual block to output the final extracted feature maps; The feature map extracted by the feature extraction module is input into the region generation module, and the region generation module adopts the RPN network to output the candidate box feature map; the candidate box feature map is unified in size by the ROI Align layer and input into the parallel Class&Box branch and Mask branch respectively, and the Class&Box branch is used to obtain the position and category information of the target object in the image, and the Mask branch is used to obtain the segmentation information of the target object in the image.

5. A method for grading and sorting coal gangue based on LIBS and image analysis as claimed in claim 4, characterized in that: The loss function of the Mask RCNN network model includes classification error loss, detection error loss, IOU error loss and improved segmentation error loss; The classification error loss and the detection error loss are both calculated using the cross entropy loss function; the IoU error loss is used to evaluate the degree of overlap between the detection box and the real object; the improved segmentation error loss is calculated using the average binary cross entropy loss function and the boundary weighted loss function.

6. A method for grading and sorting coal gangue based on LIBS and image analysis as claimed in claim 5, characterized in that: The calculation formula of the IoU error loss is: ； in, B and denote the true bounding box and the predicted bounding box respectively, Intersection represents the area of ​​the intersection region, Union represents the area of ​​the union region; The calculation formula of the improved segmentation error loss is: ； in, represents the average binary cross entropy loss function, represents the boundary weighted loss function, is the weight coefficient, which is used to adjust the boundary of the segmentation result B The extent of its impact on the overall loss; R For all segmented areas; Indicates the distance from each pixel to the boundary; y is the real label after binarization; is the predicted segmentation result after binarization, that is, the predicted label; p is the target location.

7. A coal gangue grading and sorting system based on LIBS and image analysis, characterized in that: include: A spectral data acquisition module, used to acquire spectral data of each sample in a mixed coal flow including coal lumps and coal gangue using LIBS technology; A coal block and coal gangue preliminary identification module is used to compare the acquired spectral data with the pre-established coal block spectral database, and preliminarily identify the coal block and coal gangue according to the comparison results; The gangue preliminary sorting module is used to screen the preliminarily identified coal blocks and separate the mixed coal flow, completing the preliminary sorting of the gangue in the mixed coal flow; The gangue detection module is used to obtain the visible light image of the mixed coal flow after separation, and preprocess the image. The preprocessed image is input into the gangue detection model based on the improved Mask RCNN to identify and segment all the gangue in the image and obtain the equivalent particle size of each gangue. The gangue grading and sorting module is used to divide the gangue into different levels according to the preset particle size range and to grade and sort the gangue in the mixed coal flow.

8. A coal gangue grading and sorting system based on LIBS and image analysis as claimed in claim 7, characterized in that: The acquired spectral data is compared with the pre-established coal block spectral database, including: The acquired sample spectrum data is compared with the spectrum data in the coal block spectrum database one by one according to the channels, and the difference between the sample spectrum data and the coal block spectrum data in each channel is calculated; If the difference of the spectral data of each channel of the sample is not less than the set threshold corresponding to the current channel, the sample is determined to be a coal block; On the contrary, if the difference of the spectral intensity of a certain channel in the sample is less than the set threshold corresponding to the current channel, the sample is determined to be coal gangue.

9. An electronic device, characterized in that: The invention comprises a memory and a processor and computer instructions stored in the memory and executed on the processor. When the computer instructions are executed by the processor, the steps of a coal gangue grading and sorting method based on LIBS and image analysis as described in any one of claims 1 to 6 are completed.

10. A computer-readable storage medium, characterized in that: Used to store computer instructions, which, when executed by a processor, complete the steps of a coal gangue grading and sorting method based on LIBS and image analysis as described in any one of claims 1-6.

Images