Intelligent coal gangue sorting and removing method and system

Abstract

The invention provides an intelligent coal gangue sorting and removing method and system, and relates to the technical field of coal sorting, and the method comprises the following steps: carrying out the semantic segmentation of an obtained coal flow image, and obtaining a coal region; calculating a real hue value of each pixel point in the image based on the illumination distribution function, and identifying and marking coal gangue in the coal flow image through a preset classification model; performing feature tensor analysis by using the neural network model, and determining position information of the coal gangue according to an analysis result; and rejecting the identified coal gangue from the coal flow by using a coal gangue rejecting device. According to the method, automatic recognition and removal of the coal gangue are achieved, manual intervention is reduced, the sorting efficiency is remarkably improved, feature extraction and position determination are conducted through the neural network model, accurate removal is conducted through the coal gangue removal device, and automation and high efficiency of coal sorting are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of coal sorting, and in particular to a method and system for intelligently sorting and removing coal gangue. Background Art

[0002] Gangue is a solid waste generated during coal mining and washing. It has a low carbon content, is hard and black-gray in color, and is associated with coal seams. In the coal production process, the sorting of gangue is crucial to improving the quality of coal. Given the significant differences in composition between coal and gangue, they also show obvious differences in appearance characteristics such as gloss and texture. Current technology uses these differences to accurately identify gangue through characteristics such as color, texture and gloss. Once the gangue mixed in the coal is successfully identified and marked with an anchor frame, the robotic arm will be activated to accurately grab and separate the gangue, thereby efficiently completing the sorting of the gangue.

[0003] However, despite the significant differences in appearance between coal and gangue, in the actual coal flow environment, these differences become difficult to discern due to changes in lighting, occlusion, and mixing of the two. Current technology lacks effective methods to improve this distinction, making accurate separation of coal and gangue a challenge. The instability of lighting, mutual occlusion between gangue and coal, and the mixing state between them all increase the difficulty of separation, resulting in poor sorting results, affecting the quality of coal and subsequent utilization.

[0004] Currently, no effective solution has been proposed for the problems in the related technologies. Summary of the invention

[0005] In view of this, the present invention provides a method and system for intelligently sorting and removing coal gangue to solve the above-mentioned problems.

[0006] In order to solve the above problems, the specific technical solutions adopted by the present invention are as follows: According to one aspect of the present invention, a method for intelligently sorting and removing coal gangue is provided, the method comprising the following steps: S1. Use the deep learning model to perform semantic segmentation on the acquired coal flow image to obtain the coal area; S2. For the coal flow image in the coal area, the true hue value of each pixel in the image is calculated based on the illumination distribution function, and the gangue in the coal flow image is identified and marked through a pre-set classification model; S3, based on the marked coal flow image, using the neural network model to perform feature tensor analysis, and determine the location information of the coal gangue according to the analysis results; S4. Based on the position information of the gangue, the identified gangue is removed from the coal flow using a gangue removal device.

[0007] Preferably, for a coal flow image of a coal area, calculating the true hue value of each pixel in the image based on the illumination distribution function, and identifying and marking the gangue in the coal flow image through a pre-set classification model includes the following steps: S21, converting the coal flow image of the coal region from the RGB color space to the HSV color space using a color space conversion method; S22, calculating the brightness value of each pixel in the coal flow image of the coal area in the HSV color space, and screening out connected areas from the coal flow image of the coal area; S23, constructing an illumination distribution function of the coal flow image of the coal region based on the screened connected regions, and calculating a true hue value corresponding to each pixel in the coal flow image of the coal region; S24. The true hue value corresponding to each pixel point is used as an input feature and input into a pre-set classification model to identify and mark the coal gangue in the coal flow image.

[0008] Preferably, in the HSV color space, calculating the brightness value of each pixel in the coal flow image of the coal area, and screening out the connected area from the coal flow image of the coal area comprises the following steps: S221, in the HSV color space, using the image processing library to extract the brightness value of each pixel in the coal flow image of the coal area; S222: According to the brightness value of each pixel, a connected region detection algorithm is used to select connected regions that meet a preset brightness threshold.

[0009] Preferably, constructing the illumination distribution function of the coal flow image of the coal area based on the screened connected area, and calculating the true hue value corresponding to each pixel in the coal flow image of the coal area includes the following steps: S231, for each screened connected region, calculating a connected region feature vector using the brightness value of each pixel point in the connected region; S232, establishing a relationship between the image distribution function and the true hue distribution function according to the connected region feature vector, and expanding outward with the center of the connected region as the center of the circle to obtain an extended region; S233, obtaining the brightness value of each pixel in the expansion area, and calculating the expansion area feature vector of the expansion area; S234. Constructing the illumination distribution function of the coal flow image of the coal region according to the connected region feature vector and the extended region feature vector, and calculating the true hue value corresponding to each pixel point in the coal flow image of the coal region.

[0010] Preferably, the connected region feature vector and the extended region feature vector both include the first-order moment and the second-order moment of the pixel hue.

[0011] Preferably, constructing the illumination distribution function of the coal flow image of the coal area according to the connected area feature vector and the extended area feature vector, and calculating the true hue value corresponding to each pixel in the coal flow image of the coal area includes the following steps: S2341, establishing an objective function based on minimizing the difference between the connected region feature vector and the extended region feature vector; S2342, using a Gaussian function to set an initial form of the illumination distribution function, and adjusting the parameters of the illumination distribution function through a genetic algorithm to obtain a final illumination distribution function of the coal flow image of the coal region; S2343. Calculate the true hue value of each pixel in the coal flow image of the coal area according to the final illumination distribution function of the coal flow image of the coal area.

[0012] Preferably, based on the marked coal flow image, a neural network model is used to perform feature tensor analysis, and determining the location information of coal gangue according to the analysis result includes the following steps: S31, taking the labeled coal flow image as input and inputting it into the neural network model; S32, using the grid processing layer, multiple convolutions, pooling layers and fully connected layers in the neural network model to extract features and obtain a feature tensor; S33. Process the obtained feature tensor, identify specific features related to coal gangue in the feature tensor, and determine the position information of the coal gangue in the coal flow image based on the processed feature tensor.

[0013] Preferably, the feature extraction is performed using a grid processing layer, multiple convolutions, a pooling layer, and a fully connected layer in a neural network model to obtain a feature tensor, including the following steps: S321, dividing the marked coal flow image into a plurality of grid images, and configuring a plurality of prediction frames for each grid image; S322, using the convolution layer in the neural network model to perform a convolution operation on the divided grid image, extracting local features in the grid image, and obtaining a feature map; S323. Use the pooling layer to reduce the dimension of the feature map, and convert the feature map after the dimension reduction process into a feature tensor through the fully connected layer.

[0014] Preferably, processing the obtained feature tensor, identifying specific features related to gangue in the feature tensor, and determining the position information of the gangue in the coal flow image comprises the following steps: S331, extracting the category probability of each grid from the feature tensor, and screening out candidate grids whose category probabilities are higher than the category threshold and are related to coal gangue according to a preset category threshold; S332, for each selected candidate grid, extract the corresponding predicted bounding box coordinates; S333, using a non-maximum suppression algorithm to remove candidate borders that do not meet preset conditions, to obtain a final output border; S334, extracting the position information of the gangue in the coal flow image from the final output frame.

[0015] According to another aspect of the present invention, a coal gangue intelligent sorting and removal system is provided, the system comprising: an image segmentation processing module, a coal gangue marking module, a position determination module and a coal gangue removal module, and the image segmentation processing module, the coal gangue marking module, the position determination module and the coal gangue removal module are connected in sequence; An image segmentation processing module is used to perform semantic segmentation processing on the acquired coal flow image using a deep learning model to obtain the coal area; The gangue marking module calculates the true hue value of each pixel in the coal flow image of the coal area based on the illumination distribution function, and identifies and marks the gangue in the coal flow image through a pre-set classification model; A location determination module is used to perform feature tensor analysis based on the marked coal flow image using a neural network model, and determine the location information of the coal gangue according to the analysis results; The gangue removal module is used to remove the identified gangue from the coal flow based on the position information of the gangue using the gangue removal device.

[0016] The beneficial effects of the present invention are: 1. The present invention performs semantic segmentation on coal flow images through a deep learning model, which can accurately distinguish coal areas and non-coal areas, provide an accurate basis for subsequent gangue identification, calculate the true hue value in the HSV color space, and combine it with the classification model for identification, effectively reducing the impact of illumination changes on recognition accuracy and improving the recognition accuracy of gangue.

[0017] 2. The present invention takes into account the unevenness of illumination distribution in coal flow images, and calculates the true hue value by constructing an illumination distribution function, thereby enhancing the algorithm's adaptability to illumination changes. By using a neural network model for feature tensor analysis, the complex features of coal gangue can be extracted, making the algorithm's recognition of coal gangue more robust.

[0018] 3. The present invention realizes the automatic identification and removal of coal gangue, reduces manual intervention, and significantly improves the sorting efficiency. It extracts features and determines positions through a neural network model, and uses a coal gangue removal device for precise removal, thereby realizing automation and high efficiency of coal sorting. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. In the drawings: Figure 1 This is a flow chart of a method for intelligently sorting and removing coal gangue according to an embodiment of the present invention. Figure 2 The present invention is a block diagram of a system for intelligently sorting and removing gangue according to an embodiment of the present invention.

[0020] In the figure: 1. Image segmentation and processing module; 2. Gangue marking module; 3. Position determination module; 4. Gangue removal module. DETAILED DESCRIPTION

[0021] In order to enable those skilled in the art to better understand the technical solutions in this application, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the drawings in the embodiments of this application. Obviously, the described embodiments are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of this application.

[0022] According to an embodiment of the present invention, a method and system for intelligently sorting and removing coal gangue are provided.

[0023] The present invention will now be further described with reference to the accompanying drawings and specific embodiments. Figure 1 As shown, according to one embodiment of the present invention, a method for intelligent sorting and removing of coal gangue is provided, and the method comprises the following steps: S1. Use the deep learning model to perform semantic segmentation on the acquired coal flow image to obtain the coal area; It should be noted that the purpose of using the deep learning model to perform semantic segmentation on the coal flow image is to classify each pixel in the image as belonging to the coal area or the non-coal area, which specifically includes the following steps: A large number of coal flow images are collected as training data, and these images are annotated, that is, each pixel is labeled (coal / non-coal), and the data set is divided into a training set, a validation set, and a test set.

[0024] Use labeled image data to train deep learning models, including FCN (Fully Convolutional Network), U-Net, SegNet, DeepLab, etc. During the training process, the model will learn the characteristics of coal and non-coal areas and gradually optimize its segmentation performance. By adjusting model parameters, optimizers, loss functions, etc., the segmentation accuracy and efficiency of the model can be improved. The pre-acquired coal flow image is input into the trained deep learning model, which classifies the image pixel by pixel and generates a segmentation result image of the same size as the input image. In the segmentation result image, each pixel is labeled as coal or non-coal.

[0025] S2. For the coal flow image in the coal area, the true hue value of each pixel in the image is calculated based on the illumination distribution function, and the gangue in the coal flow image is identified and marked through a pre-set classification model; As a preferred embodiment, for a coal flow image of a coal area, the true hue value of each pixel in the image is calculated based on the illumination distribution function, and the gangue in the coal flow image is identified and marked by a pre-set classification model, including the following steps: S21, converting the coal flow image of the coal region from the RGB color space to the HSV color space using a color space conversion method; Specifically, the color space conversion method is used to convert the coal flow image of the coal area from the RGB color space to the HSV color space. The maximum and minimum RGB component values ​​of each pixel are calculated to determine the brightness (Value) and saturation (Saturation) in the HSV color space, and the hue (Hue) is calculated according to the channel corresponding to the maximum component value and the minimum component value, thereby realizing the conversion of the color space.

[0026] S22, calculating the brightness value of each pixel in the coal flow image of the coal area in the HSV color space, and screening out connected areas from the coal flow image of the coal area; As a preferred implementation, in the HSV color space, calculating the brightness value of each pixel in the coal flow image of the coal area, and screening out the connected area from the coal flow image of the coal area includes the following steps: S221, in the HSV color space, using the image processing library to extract the brightness value of each pixel in the coal flow image of the coal area; It should be noted that in the HSV color space, the brightness value directly corresponds to the brightness information of the image. The brightness channel is separated from the HSV image using the channel separation function in the image processing library. The separated brightness channel image contains the brightness value of each pixel. These values ​​are usually in the range of 0 to 255 and can be used directly as brightness values.

[0027] S222: According to the brightness value of each pixel, a connected region detection algorithm is used to select connected regions that meet a preset brightness threshold.

[0028] Specifically, according to the characteristics of the coal area, a suitable brightness threshold is set. This threshold can be used to distinguish between coal and non-coal areas. The brightness image (i.e., the brightness channel image extracted previously) is binarized using the set brightness threshold. In the binary image, pixels with brightness values ​​higher than the threshold are set to white (or 255), indicating coal areas; pixels with brightness values ​​lower than the threshold are set to black (or 0), indicating non-coal areas. A connected domain detection algorithm is applied to the binary image. The connected domain detection algorithm can find all connected white pixel blocks (i.e., coal areas) in the image.

[0029] S23, constructing an illumination distribution function of the coal flow image of the coal region based on the screened connected regions, and calculating a true hue value corresponding to each pixel in the coal flow image of the coal region; As a preferred implementation, based on the screened connected areas, constructing the illumination distribution function of the coal flow image of the coal area, and calculating the true hue value corresponding to each pixel in the coal flow image of the coal area includes the following steps: S231, for each screened connected region, calculating a connected region feature vector using the brightness value of each pixel point in the connected region; It should be noted that feature vectors are usually used to describe certain characteristics of a region, such as color distribution, texture, etc. In this step, the feature vector includes the first-order moment (mean) and second-order moment (variance) of the pixel hue. These statistics can help understand the distribution of hue in connected regions.

[0030] S232, establishing a relationship between the image distribution function and the true hue distribution function according to the connected region feature vector, and expanding outward with the center of the connected region as the center of the circle to obtain an extended region; It should be noted that since changes in illumination may affect the observation of hue values, this effect needs to be corrected in some way. One possible approach is to use a machine learning model that can predict the true hue distribution based on the connected region feature vector. In addition, it is also possible to consider expanding outward from the center of the connected region to obtain an extended region so as to analyze changes in illumination and hue over a larger range.

[0031] S233, obtaining the brightness value of each pixel in the expansion area, and calculating the expansion area feature vector of the expansion area; As a preferred implementation, both the connected region feature vector and the extended region feature vector include the first-order moment and the second-order moment of the pixel hue.

[0032] S234. Constructing the illumination distribution function of the coal flow image of the coal region according to the connected region feature vector and the extended region feature vector, and calculating the true hue value corresponding to each pixel point in the coal flow image of the coal region.

[0033] As a preferred implementation, constructing the illumination distribution function of the coal flow image of the coal area according to the connected region feature vector and the extended region feature vector, and calculating the true hue value corresponding to each pixel in the coal flow image of the coal area includes the following steps: S2341, establishing an objective function based on minimizing the difference between the connected region feature vector and the extended region feature vector; Specifically, for each region, these statistics can be combined into a feature vector. For example, the connected region feature vector f c =[ f c1 , f c2 ,..., f cn ]; Extended region feature vector f e =[ f e1 , f e2 ,..., f en ]; The goal is to find a light distribution function I ( x , y ), so that after applying this illumination distribution function, the processed connected region feature vector is as close as possible to the extended region feature vector.

[0034] In order to calculate the difference between two feature vectors, a metric needs to be selected. Common metrics include Euclidean distance, Mahalanobis distance, cosine similarity, mean square error, etc. Here, mean square error is preferred as the metric. Then the objective function can be constructed as follows J : ; In the formula, f ci ( I ) represents the illumination distribution function I ( x , y ) after processing to obtain the connected region feature vector i Quantity; f ei ( I ) represents the illumination distribution functionI ( x , y ) after processing to obtain the extended region feature vector i Quantity; S2342, using a Gaussian function to set an initial form of the illumination distribution function, and adjusting the parameters of the illumination distribution function through a genetic algorithm to obtain a final illumination distribution function of the coal flow image of the coal region; It should be noted that the Gaussian function, as a continuous and smooth probability distribution function, can well simulate the distribution laws of many phenomena in nature, including the spatial distribution of light intensity.

[0035] Specifically, the illumination distribution function is expressed as a Gaussian function, whose parameters include mean and standard deviation. The mean represents the center position of the illumination distribution, while the standard deviation describes the degree of dispersion of the illumination distribution. These parameters can be initially set based on experience or the results of data preprocessing.

[0036] Genetic algorithm is an optimization algorithm that simulates the principles of natural selection and genetics. It continuously iterates and optimizes individuals (here, the parameters of the illumination distribution function) through operations such as selection, crossover, and mutation until the optimal solution is found or the stopping condition is met. Through the continuous iterative optimization of the genetic algorithm, the final illumination distribution function can be obtained, and its parameters have been adjusted to the optimal state, which can accurately reflect the illumination distribution in the coal flow image of the coal area.

[0037] S2343. Calculate the true hue value of each pixel in the coal flow image of the coal area according to the final illumination distribution function of the coal flow image of the coal area.

[0038] Specifically, since lighting conditions often have a significant impact on the hue of an image, the observed hue value deviates from the true hue of the coal. In order to obtain the true hue value of each pixel, the influence of lighting needs to be removed, usually by dividing the original hue value by the corresponding light intensity. Since the hue value is usually between 0 and 360 degrees, the calculated true hue value needs to be normalized to ensure that it is within the valid range.

[0039] S24. The true hue value corresponding to each pixel point is used as an input feature and input into a pre-set classification model to identify and mark the coal gangue in the coal flow image.

[0040] Specifically, common classification models include support vector machines, decision trees, random forests, etc.; the classification model can be pre-trained, and by inputting the true hue value of each pixel into the pre-trained classification model, the model will output the classification result of each pixel, that is, whether the pixel belongs to coal or coal gangue.

[0041] S3, based on the marked coal flow image, using the neural network model to perform feature tensor analysis, and determine the location information of the coal gangue according to the analysis results; Specifically, the neural network model adopts the YOLO neural network model.

[0042] As a preferred embodiment, based on the marked coal flow image, a neural network model is used to perform feature tensor analysis, and the location information of coal gangue is determined according to the analysis result, including the following steps: S31, taking the labeled coal flow image as input and inputting it into the neural network model; S32, using the grid processing layer, multiple convolutions, pooling layers and fully connected layers in the neural network model to extract features and obtain a feature tensor; As a preferred implementation, the grid processing layer, multiple convolutions, pooling layers and fully connected layers in the neural network model are used to extract features, and obtaining the feature tensor includes the following steps: S321, dividing the marked coal flow image into a plurality of grid images, and configuring a plurality of prediction frames for each grid image; It should be noted that each grid is assigned several predicted bounding boxes. These bounding boxes are used to predict and locate objects (such as coal gangue) in the image. Each bounding box has its own confidence score, which indicates the probability of the object existing in the bounding box and the accuracy of the bounding box.

[0043] S322, using the convolution layer in the neural network model to perform a convolution operation on the divided grid image, extracting local features in the grid image, and obtaining a feature map; It should be noted that the convolution layer uses a series of filters (or convolution kernels) that slide over the image to extract features at different locations, such as edges, color blocks, textures, etc. The output is called a feature map, which is a compressed representation of the original input image that highlights the most critical visual features.

[0044] S323. Use the pooling layer to reduce the dimension of the feature map, and convert the feature map after the dimension reduction process into a feature tensor through the fully connected layer.

[0045] S33. Process the obtained feature tensor, identify specific features related to coal gangue in the feature tensor, and determine the position information of the coal gangue in the coal flow image based on the processed feature tensor.

[0046] As a preferred implementation, processing the obtained feature tensor, identifying specific features related to gangue in the feature tensor, and determining the position information of the gangue in the coal flow image includes the following steps: S331, extracting the category probability of each grid from the feature tensor, and screening out candidate grids whose category probabilities are higher than the category threshold and are related to coal gangue according to a preset category threshold; It should be noted that the neural network predicts a series of class probabilities for each grid in the feature tensor. These probabilities indicate the possibility that the grid contains a specific object (in this case, coal gangue). By setting a preset class threshold, grids with class probabilities higher than this threshold can be screened out, and these grids are considered candidate grids containing coal gangue.

[0047] S332, for each selected candidate grid, extract the corresponding predicted bounding box coordinates; It should be noted that each frame includes its position in the image (eg, the coordinates of the four corners of the rectangular frame). These frames are predicted by the neural network in step S321 and represent specific locations where coal gangue may exist.

[0048] S333, using a non-maximum suppression algorithm to remove candidate borders that do not meet preset conditions, to obtain a final output border; It should be noted that non-maximum suppression is a commonly used algorithm for removing overlapping borders in object detection tasks. If multiple borders overlap and point to the same object, the border with the highest class probability will be retained, and the other overlapping borders will be removed.

[0049] S334, extracting the position information of the gangue in the coal flow image from the final output frame.

[0050] S4. Based on the position information of the gangue, the identified gangue is removed from the coal flow using a gangue removal device.

[0051] Specifically, according to the location information of the gangue, a suitable gangue removal device is designed. Common removal devices include pneumatic manipulators, which can accurately grasp and remove the gangue according to the location information of the gangue.

[0052] like Figure 2 As shown, according to another embodiment of the present invention, a coal gangue intelligent sorting and removal system is provided, the system: image segmentation processing module 1, coal gangue marking module 2, position determination module 3 and coal gangue removal module 4, and the image segmentation processing module, coal gangue marking module, position determination module and coal gangue removal module are connected in sequence; The image segmentation processing module 1 is used to perform semantic segmentation processing on the acquired coal flow image using a deep learning model to obtain a coal area; The gangue marking module 2 calculates the true hue value of each pixel in the coal flow image of the coal area based on the illumination distribution function, and identifies and marks the gangue in the coal flow image through a pre-set classification model; The position determination module 3 is used to perform feature tensor analysis based on the marked coal flow image using a neural network model, and determine the position information of the coal gangue according to the analysis result; The gangue removal module 4 is used to remove the identified gangue from the coal flow by using the gangue removal device based on the position information of the gangue.

[0053] It should be understood by those skilled in the art that the embodiments of the present invention may be provided as methods, systems or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware aspects. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, optical storage, etc.) containing computer-usable program code.

[0054] The specific embodiments described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for intelligent sorting and removal of coal gangue, characterized in that: The method comprises the following steps: S1. Use the deep learning model to perform semantic segmentation on the acquired coal flow image to obtain the coal area; S2. For the coal flow image in the coal area, the true hue value of each pixel in the image is calculated based on the illumination distribution function, and the gangue in the coal flow image is identified and marked through a pre-set classification model; S3, based on the marked coal flow image, using the neural network model to perform feature tensor analysis, and determine the location information of the coal gangue according to the analysis results; S4. Based on the position information of the gangue, the identified gangue is removed from the coal flow using a gangue removal device.

2. The method for intelligent sorting and removing coal gangue according to claim 1 is characterized in that: The method of calculating the true hue value of each pixel in the coal flow image of the coal area based on the illumination distribution function and identifying and marking the gangue in the coal flow image by a pre-set classification model includes the following steps: S21, converting the coal flow image of the coal region from the RGB color space to the HSV color space using a color space conversion method; S22, calculating the brightness value of each pixel in the coal flow image of the coal area in the HSV color space, and screening out connected areas from the coal flow image of the coal area; S23, constructing an illumination distribution function of the coal flow image of the coal region based on the screened connected regions, and calculating a true hue value corresponding to each pixel in the coal flow image of the coal region; S24. The true hue value corresponding to each pixel point is used as an input feature and input into a pre-set classification model to identify and mark the coal gangue in the coal flow image.

3. The method for intelligent sorting and removing coal gangue according to claim 2 is characterized in that: The method of calculating the brightness value of each pixel in the coal flow image of the coal area in the HSV color space and screening out the connected area from the coal flow image of the coal area includes the following steps: S221, in the HSV color space, using the image processing library to extract the brightness value of each pixel in the coal flow image of the coal area; S222: According to the brightness value of each pixel, a connected region detection algorithm is used to select connected regions that meet a preset brightness threshold.

4. The method for intelligently sorting and removing coal gangue according to claim 2, characterized in that: The method of constructing the illumination distribution function of the coal flow image of the coal area based on the screened connected area and calculating the true hue value corresponding to each pixel in the coal flow image of the coal area includes the following steps: S231, for each screened connected region, calculating a connected region feature vector using the brightness value of each pixel in the connected region; S232, establishing a relationship between the image distribution function and the true hue distribution function according to the connected region feature vector, and expanding outward with the center of the connected region as the center of the circle to obtain an extended region; S233, obtaining the brightness value of each pixel in the expansion area, and calculating the expansion area feature vector of the expansion area; S234. Constructing the illumination distribution function of the coal flow image of the coal region according to the connected region feature vector and the extended region feature vector, and calculating the true hue value corresponding to each pixel point in the coal flow image of the coal region.

5. The method for intelligently sorting and removing coal gangue according to claim 4, characterized in that: The connected region feature vector and the extended region feature vector both include the first-order moment and the second-order moment of the pixel hue.

6. The method for intelligently sorting and removing coal gangue according to claim 4, characterized in that: The method of constructing the illumination distribution function of the coal flow image of the coal area according to the connected area feature vector and the extended area feature vector, and calculating the true hue value corresponding to each pixel in the coal flow image of the coal area includes the following steps: S2341, establishing an objective function based on minimizing the difference between the connected region feature vector and the extended region feature vector; S2342, using a Gaussian function to set an initial form of the illumination distribution function, and adjusting the parameters of the illumination distribution function through a genetic algorithm to obtain a final illumination distribution function of the coal flow image of the coal region; S2343. Calculate the true hue value of each pixel in the coal flow image of the coal area according to the final illumination distribution function of the coal flow image of the coal area.

7. The method for intelligently sorting and removing coal gangue according to claim 1, characterized in that: The method of performing feature tensor analysis based on the marked coal flow image using a neural network model and determining the location information of coal gangue according to the analysis result includes the following steps: S31, taking the labeled coal flow image as input and inputting it into the neural network model; S32, using the grid processing layer, multiple convolutions, pooling layers and fully connected layers in the neural network model to extract features and obtain a feature tensor; S33. Process the obtained feature tensor, identify specific features related to coal gangue in the feature tensor, and determine the position information of the coal gangue in the coal flow image based on the processed feature tensor.

8. The method for intelligently sorting and removing coal gangue according to claim 1, characterized in that: The method of extracting features by using the grid processing layer, multiple convolutions, pooling layers and fully connected layers in the neural network model to obtain a feature tensor includes the following steps: S321, dividing the marked coal flow image into a plurality of grid images, and configuring a plurality of prediction frames for each grid image; S322, using the convolution layer in the neural network model to perform a convolution operation on the divided grid image, extracting local features in the grid image, and obtaining a feature map; S323. Use the pooling layer to reduce the dimension of the feature map, and convert the feature map after the dimension reduction process into a feature tensor through the fully connected layer.

9. The method for intelligently sorting and removing coal gangue according to claim 8, characterized in that: The processing of the obtained feature tensor, identifying the specific features related to the gangue in the feature tensor, and determining the position information of the gangue in the coal flow image includes the following steps: S331, extracting the category probability of each grid from the feature tensor, and screening out candidate grids whose category probabilities are higher than the category threshold and are related to coal gangue according to a preset category threshold; S332, for each selected candidate grid, extract the corresponding predicted bounding box coordinates; S333, using a non-maximum suppression algorithm to remove candidate borders that do not meet preset conditions, to obtain a final output border; S334, extracting the position information of the gangue in the coal flow image from the final output frame.

10. A coal gangue intelligent sorting and removal system, used to implement the coal gangue intelligent sorting and removal method according to any one of claims 1 to 9, characterized in that: The system includes an image segmentation processing module, a gangue marking module, a position determination module and a gangue removal module, and the image segmentation processing module, the gangue marking module, the position determination module and the gangue removal module are connected in sequence; The image segmentation processing module is used to perform semantic segmentation processing on the acquired coal flow image using a deep learning model to obtain a coal area; The gangue marking module calculates the true hue value of each pixel in the coal flow image of the coal area based on the illumination distribution function, and identifies and marks the gangue in the coal flow image through a pre-set classification model; The position determination module is used to perform feature tensor analysis based on the marked coal flow image using a neural network model, and determine the position information of the coal gangue according to the analysis result; The gangue removal module is used to remove the identified gangue from the coal flow by using the gangue removal device based on the position information of the gangue.

Images