Recognition Method for Coal Flow Blockage State of Scraper Conveyor in Fully Mechanized Coal Mining Face

Abstract

The present disclosure proposes a method for identifying the coal flow blockage state of a scraper conveyor in a fully mechanized coal mining face. The method includes: constructing an initial coal flow blockage state recognition model, where the initial coal flow blockage state recognition model is constructed based on the YOLOv8n object detection algorithm, the StarNet backbone network, the context star fusion module, and the weight-sharing detection head, and the StarNet backbone network serves as the backbone of the YOLOv8n object detection algorithm; obtaining a target data set, where the target data set includes multiple sample coal flow images, and the sample coal flow images contain scraper conveyor coal flow blockage information; training the initial coal flow blockage state recognition model based on the target data set to obtain a target coal flow blockage state recognition model; and identifying the coal flow blockage state of the target scraper conveyor based on the target coal flow blockage state recognition model. Thus, by optimizing the model structure and introducing innovative modules, the accuracy and real-time performance of the coal flow blockage detection method can be effectively improved.

Description

Technical Field

[0001] The present disclosure relates to the field of image recognition technology, and particularly to a method for identifying the coal flow blockage state of a scraper conveyor in a fully mechanized coal mining face of a coal mine. Background Art

[0002] Coal flow refers to the flowing state of coal on a conveyor belt under the action of equipment such as a scraper conveyor. Generally, coal flow is composed of a large number of smaller coal particles or coal blocks, and these coal particles are relatively loose during the conveying process and can move on the conveyor belt like a fluid. Coal flow blockage is a common safety hazard in the fully mechanized coal mining face of a coal mine. Especially during the mining process, abnormal blockage of the coal flow may cause the scraper conveyor to break, the shearer to malfunction, poor ventilation, and even lead to more serious mine accidents.

[0003] In the related art, when identifying the coal flow blockage state of a scraper conveyor, the detection method has low accuracy and poor real-time performance. Summary of the Invention

[0004] The present disclosure aims to solve at least one of the technical problems in the related art to some extent.

[0005] To this end, the purpose of the present disclosure is to propose a method for identifying the coal flow blockage state of a scraper conveyor in a fully mechanized coal mining face of a coal mine, which can effectively improve the accuracy and real-time performance of the coal flow blockage detection method by optimizing the model structure and introducing innovative modules.

[0006] To achieve the above object, the method for identifying the coal flow blockage state of a scraper conveyor in a fully mechanized coal mining face of a coal mine proposed by an embodiment of the present disclosure includes:

[0007] Construct an initial coal flow blockage state recognition model, wherein the initial coal flow blockage state recognition model is constructed based on the YOLOv8n object detection algorithm, the StarNet backbone network, the context star fusion module, and the weight-sharing detection head, and the StarNet backbone network serves as the backbone of the YOLOv8n object detection algorithm;

[0008] Obtain a target data set, wherein the target data set includes a plurality of sample coal flow images, and the sample coal flow images contain scraper conveyor coal flow blockage information;

[0009] Train the initial coal flow blockage state recognition model based on the target data set to obtain a target coal flow blockage state recognition model;

[0010] Identify the coal flow blockage state of the target scraper conveyor based on the target coal flow blockage state recognition model.

[0011] The method for identifying the coal flow blockage state of the scraper conveyor in the fully mechanized coal mining face provided by the present disclosure constructs an initial coal flow blockage state recognition model. Among them, the initial coal flow blockage state recognition model is constructed based on the YOLOv8n object detection algorithm, the StarNet backbone network, the context star fusion module, and the weight-sharing detection head. The StarNet backbone network serves as the backbone of the YOLOv8n object detection algorithm; obtain a target data set, where the target data set includes multiple sample coal flow images, and the sample coal flow images contain information on the coal flow blockage of the scraper conveyor; train the initial coal flow blockage state recognition model based on the target data set to obtain a target coal flow blockage state recognition model; identify the coal flow blockage state of the target scraper conveyor based on the target coal flow blockage state recognition model. Thus, by optimizing the model structure and introducing innovative modules, the accuracy and real-time performance of the coal flow blockage detection method can be effectively improved.

[0012] Additional aspects and advantages of the present disclosure will be given in part in the following description, become apparent in part from the following description, or be understood through the practice of the present disclosure. Brief Description of the Drawings

[0013] The above-mentioned and / or additional aspects and advantages of the present disclosure will become apparent and easy to understand from the following description of the embodiments in conjunction with the drawings, where:

[0014] Figure 1 is a schematic flowchart of the method for identifying the coal flow blockage state of the scraper conveyor in the fully mechanized coal mining face proposed in an embodiment of the present disclosure;

[0015] Figure 2 is a schematic flowchart of the method for identifying the coal flow blockage state of the scraper conveyor in the fully mechanized coal mining face proposed in another embodiment of the present disclosure;

[0016] Figure 3 is a schematic flowchart of the method for identifying the coal flow blockage state of the scraper conveyor in the fully mechanized coal mining face proposed in another embodiment of the present disclosure;

[0017] Figure 4 is a schematic diagram of the principle of the method for identifying the coal flow blockage state of the scraper conveyor in the fully mechanized coal mining face proposed according to the present disclosure;

[0018] Figure 5 is a schematic diagram of the network structure of the object detection method proposed according to the present disclosure;

[0019] Figure 6 is a schematic diagram of the context anchor point attention mechanism module proposed according to the present disclosure. Detailed Embodiments

[0020] Embodiments of the present disclosure will be described in detail below. Examples of the embodiments are shown in the accompanying drawings, where like or similar reference numerals denote like or similar elements or elements having like or similar functions throughout. The embodiments described below by referring to the accompanying drawings are exemplary and are only used to explain the present disclosure and should not be construed as a limitation of the present disclosure. On the contrary, the embodiments of the present disclosure include all variations, modifications, and equivalents falling within the spirit and scope of the appended claims.

[0021] Figure 1 It is a schematic flow chart of a method for identifying the coal flow blockage state of a shearer in a fully mechanized coal mining face proposed in an embodiment of the present disclosure.

[0022] It should be noted that the execution subject of the method for identifying the coal flow blockage state of the shearer in the fully mechanized coal mining face in this embodiment is a device for identifying the coal flow blockage state of the shearer in the fully mechanized coal mining face, and this device can be implemented in software and / or hardware.

[0023] As Figure 1 shown, the method for identifying the coal flow blockage state of the shearer in the fully mechanized coal mining face includes:

[0024] S101: Construct an initial coal flow blockage state recognition model. Among them, the initial coal flow blockage state recognition model is constructed based on the YOLOv8n object detection algorithm, the StarNet backbone network, the contextual star fusion module, and the weight-sharing detection head. The StarNet backbone network serves as the backbone of the YOLOv8n object detection algorithm.

[0025] Among them, the initial coal flow blockage state recognition model refers to a model pre-constructed in the embodiments of the present disclosure for identifying the coal flow blockage state. This initial coal flow blockage state recognition model can be constructed based on the YOLOv8n object detection algorithm, using the StarNet backbone network as the backbone network of the YOLOv8n object detection algorithm, and adding a contextual star fusion module and a weight-sharing detection head.

[0026] Among them, the contextual star fusion module (CSFM) is an innovative design of StarNet in object detection tasks, aiming to enhance the model's global understanding of objects in images. The main function of this module is to effectively fuse local information and global context information to improve detection performance.

[0027] Among them, the Weight Shared Detection Head (WSD Head) is a design method in the object detection model, aiming to improve the parameter efficiency of the detection head part and reduce the computational overhead. Its core idea is to reduce the redundant parameters in the model by sharing the weights of multiple detection heads, while still being able to perform efficient detection on multiple tasks or multiple different object categories.

[0028] In the embodiments of the present disclosure, when constructing the initial coal flow blockage state recognition model, it can provide a reliable execution tool for subsequent coal flow blockage state recognition.

[0029] S102: Obtain a target data set, where the target data set includes multiple sample coal flow images, and the sample coal flow images contain scraper conveyor coal flow blockage information.

[0030] Among them, the target data set can refer to the data set obtained in the embodiments of the present disclosure for training the initial coal flow blockage state recognition model.

[0031] In the embodiments of the present disclosure, when obtaining the target data set, it can be to collect the coal transportation video of the scraper conveyor by using the working face camera, perform frame extraction processing to obtain the first data set, and merge the first data set with the public data set to obtain the target data set.

[0032] Among them, the sample coal flow image can refer to an image containing information related to the scraper conveyor coal flow blockage.

[0033] That is to say, in the embodiments of the present disclosure, an image containing coal flow blockage information can be obtained to construct the target data set, thereby providing reliable data support for subsequent training of the initial coal flow blockage state recognition model.

[0034] S103: Train the initial coal flow blockage state recognition model based on the target data set to obtain the target coal flow blockage state recognition model.

[0035] Among them, the target coal flow blockage state recognition model refers to the coal flow blockage state recognition model obtained by training the initial coal flow blockage state recognition model via the target data set.

[0036] In the embodiments of the present disclosure, when training the initial coal flow blockage state recognition model based on the target data set, the target data set can be divided into a training set, a validation set, and a test set according to a certain ratio, and then the processed data set is input into the initial coal flow blockage state recognition model for training and the model weight file with the highest accuracy is saved.

[0037] In the embodiments of the present disclosure, when training an initial coal flow blockage state recognition model based on a target data set to obtain a target coal flow blockage state recognition model, it can provide a reliable execution tool for subsequent recognition of the coal flow blockage state of the target scraper conveyor.

[0038] S104: Recognize the coal flow blockage state of the target scraper conveyor based on the target coal flow blockage state recognition model.

[0039] The target scraper conveyor herein refers to the scraper conveyor for which the coal flow blockage state is to be recognized in the embodiments of the present disclosure.

[0040] That is to say, in the embodiments of the present disclosure, after training an initial coal flow blockage state recognition model based on a target data set to obtain a target coal flow blockage state recognition model, the coal flow blockage state of the target scraper conveyor can be recognized based on the target coal flow blockage state recognition model, thereby effectively improving the recognition efficiency and intelligence level of the coal flow blockage state.

[0041] In this embodiment, an initial coal flow blockage state recognition model is constructed. Among them, the initial coal flow blockage state recognition model is constructed based on the YOLOv8n object detection algorithm, the StarNet backbone network, the context star fusion module, and the weight-sharing detection head. The StarNet backbone network serves as the backbone of the YOLOv8n object detection algorithm; a target data set is obtained, where the target data set includes multiple sample coal flow images, and the sample coal flow images contain scraper conveyor coal flow blockage information; the initial coal flow blockage state recognition model is trained based on the target data set to obtain a target coal flow blockage state recognition model; the coal flow blockage state of the target scraper conveyor is recognized based on the target coal flow blockage state recognition model. Thus, the accuracy and real-time performance of the coal flow blockage detection method can be effectively improved by optimizing the model structure and introducing innovative modules.

[0042] Figure 2 It is a schematic flowchart of a method for recognizing the coal flow blockage state of a scraper conveyor in a fully mechanized coal mining face proposed in another embodiment of the present disclosure.

[0043] As Figure 2 shown, the method for recognizing the coal flow blockage state of the scraper conveyor in the fully mechanized coal mining face includes:

[0044] S201: Construct an initial coal flow blockage state recognition model, where the initial coal flow blockage state recognition model is constructed based on the YOLOv8n object detection algorithm, the StarNet backbone network, the context star fusion module, and the weight-sharing detection head, and the StarNet backbone network serves as the backbone of the YOLOv8n object detection algorithm.

[0045] S202: Obtain a target data set, where the target data set includes multiple sample coal flow images, and the sample coal flow images contain scraper conveyor coal flow blockage information.

[0046] S203: Train an initial coal flow blockage state recognition model based on a target data set to obtain a target coal flow blockage state recognition model.

[0047] For the descriptions of S201 - S203, specific reference can be made to the above - mentioned embodiments, which will not be elaborated here.

[0048] S204: Obtain real - time coal flow images of a target scraper conveyor based on a preset time interval.

[0049] The preset time interval refers to the time interval of the coal flow images corresponding to the target scraper conveyor preset in the present disclosure. For example, it can be 1 second, 10 seconds, etc., and there is no limit to this.

[0050] The real - time coal flow image refers to the coal flow image collected in the actual production operation environment.

[0051] That is to say, in the embodiments of the present disclosure, real - time coal flow images of the target scraper conveyor can be collected and obtained based on a preset time interval, thereby providing reliable data support for subsequent coal flow blockage state recognition.

[0052] S205: Perform recognition processing on the real - time coal flow images based on the target coal flow blockage state recognition model to obtain a coal flow blockage state recognition result.

[0053] The coal flow blockage state recognition result can be used to indicate whether there is a coal flow blockage in the target scraper conveyor, or can further indicate the degree of coal flow blockage.

[0054] That is to say, in the embodiments of the present disclosure, after training an initial coal flow blockage state recognition model based on a target data set to obtain a target coal flow blockage state recognition model, real - time coal flow images of the target scraper conveyor can be obtained based on a preset time interval; recognition processing is performed on the real - time coal flow images based on the target coal flow blockage state recognition model to obtain a coal flow blockage state recognition result. Thus, real - time monitoring of the coal flow state can be realized, thereby effectively improving the timeliness and reliability of the obtained coal flow blockage state recognition.

[0055] S206: When the coal flow blockage state recognition result indicates that there is a coal flow blockage state in the real - time coal flow image, generate an alarm message based on the real - time coal flow image and perform a coal flow blockage early warning based on the alarm message.

[0056] The alarm message can be used for coal flow blockage early warning. For example, the alarm message can be used to indicate whether there is a coal flow blockage in the target scraper conveyor, or can also be used to indicate the degree of coal flow blockage, etc., and there is no limit to this.

[0057] Optionally, in some embodiments, when generating the alarm information based on the real-time coal flow image, it may be to determine the volume of the coal pile where the target scraper conveyor is in a blocked state according to the real-time coal flow image, and generate the alarm information based on the volume of the coal pile. Thus, the obtained alarm information can accurately indicate the degree of blockage of the target scraper conveyor, effectively improving the practicality of the alarm information.

[0058] Optionally, in some embodiments, when generating the alarm information based on the real-time coal flow image, it may also be to determine the position of the coal pile where the target scraper conveyor is in a blocked state according to the real-time coal flow image, and generate the alarm information based on the position of the coal pile. Thus, the obtained alarm information can accurately indicate the position of the coal pile in a blocked state, facilitating the rapid positioning of the blocked coal flow, and effectively improving the efficiency of coal flow blockage handling.

[0059] Optionally, in some embodiments, when performing coal flow blockage early warning based on the alarm information, it may include at least one of the following:

[0060] Performing a light early warning based on the alarm information;

[0061] Performing a voice early warning based on the alarm information;

[0062] Generating a coal flow blockage status chart based on the alarm information, and performing a visual early warning based on the coal flow blockage status chart.

[0063] Thus, a personalized method can be adopted to implement the coal flow blockage early warning, effectively improving the early warning effect.

[0064] Among them, the coal flow blockage status chart refers to a related chart describing the coal flow blockage status. For example, it can be a line chart.

[0065] That is to say, in the embodiments of the present disclosure, when the coal flow blockage status recognition result indicates that there is a coal flow in a blocked state in the real-time coal flow image, the alarm information can be generated according to the real-time coal flow image, and the coal flow blockage early warning can be performed based on the alarm information. Thus, the corresponding alarm information can be accurately generated based on the real-time coal flow image, and the coal flow blockage early warning can be performed based on the alarm information.

[0066] In this embodiment, by obtaining the real-time coal flow image of the target scraper conveyor at preset time intervals; performing recognition processing on the real-time coal flow image based on the target coal flow blockage status recognition model to obtain the coal flow blockage status recognition result. Thus, the real-time monitoring of the coal flow state can be realized, effectively improving the timeliness and reliability of the obtained coal flow blockage status recognition. When the coal flow blockage status recognition result indicates that there is a coal flow in a blocked state in the real-time coal flow image, the alarm information can be generated according to the real-time coal flow image, and the coal flow blockage early warning can be performed based on the alarm information. Thus, the corresponding alarm information can be accurately generated based on the real-time coal flow image, and the coal flow blockage early warning can be performed based on the alarm information.

[0067] Figure 3 It is a schematic flow chart of a method for identifying the coal flow blockage state of a shearer in a fully mechanized coal mining face proposed in another embodiment of the present disclosure.

[0068] As Figure 3 shown, the method for identifying the coal flow blockage state of the shearer in the fully mechanized coal mining face includes:

[0069] S301: Construct an initial coal flow blockage state recognition model, where the initial coal flow blockage state recognition model is constructed based on the YOLOv8n object detection algorithm, the StarNet backbone network, the context star fusion module, and the weight-sharing detection head. The StarNet backbone network serves as the backbone of the YOLOv8n object detection algorithm.

[0070] S302: Obtain a target data set, where the target data set includes multiple sample coal flow images, and the sample coal flow images contain shearer coal flow blockage information.

[0071] S303: Train the initial coal flow blockage state recognition model based on the target data set to obtain a target coal flow blockage state recognition model.

[0072] S304: Obtain real-time coal flow images of the target shearer based on a preset time interval.

[0073] S305: Perform recognition processing on the real-time coal flow images based on the target coal flow blockage state recognition model to obtain a coal flow blockage state recognition result.

[0074] For the descriptions of S301 - S305, specific reference can be made to the above embodiments, which will not be elaborated here.

[0075] S306: When the coal flow blockage state recognition result indicates that there is a coal flow in a blocked state in the real-time coal flow image, send a preset control instruction to the target shearer, where the preset control instruction is used to instruct the target shearer to stop running.

[0076] Among them, the preset control instruction can refer to a pre-configured control instruction for instructing the target shearer to stop running.

[0077] In this embodiment, when the coal flow blockage state recognition result indicates that there is a coal flow in a blocked state in the real-time coal flow image, a preset control instruction is sent to the target shearer, where the preset control instruction is used to instruct the target shearer to stop running. Thus, it is possible to timely instruct the target shearer to stop running based on the preset control instruction when the coal flow is in a blocked state, so as to effectively improve the safety and robustness of the operation process of the target shearer.

[0078] Combining the above embodiments, as Figure 4 shown, Figure 4It is a schematic diagram of the principle of the method for identifying the coal flow blockage state of the scraper conveyor in the fully mechanized coal mining face proposed according to the present disclosure. Firstly, the coal transportation video of the scraper conveyor is collected by the working face camera, frame extraction is performed, and it is merged with the data set. The merged data set is manually visually labeled, and only the areas where the coal flow is in a blocked state are labeled. The training set, validation set, and test set are divided according to a certain ratio. Then, the processed data set is input into Star-YOLOv8 for training, and the model weight file with the highest accuracy is saved. Finally, the weight file, inference algorithm, and post-processing algorithm are deployed to the back-end service of the mine safety inspection software, and the service is deployed to the Linux system server. The real-time video stream provided by the algorithm bound by the configuration file and the working face camera is sent to the back-end service. It can also be deployed to an edge computing server. Each frame of data is processed in real time. If the coal flow is blocked, an immediate response instruction is sent to the PLC. The PLC then sends an instruction to control the relay to light up, and the relay sends an instruction message to control the buzzer to give an alarm warning, and the operation of the scraper conveyor can also be shut down. The coal flow blockage position, pictures, and video clips are uploaded to the server for storage, and the video inference stream is displayed in real time on the front end of the mine safety inspection software, and the alarm information is displayed in real time in a visual form such as a chart on the early warning interface.

[0079] The present disclosure proposes a lightweight object detection method (Star-YOLOv8) based on YOLOv8n to improve the computational efficiency and detection accuracy of the model. Figure 5 As shown Figure 5 It is a schematic diagram of the network structure of the object detection method proposed according to the present disclosure.

[0080] Firstly, the present disclosure replaces the backbone network of YOLOv8 with StarNet. StarNet is an efficient neural network architecture that performs feature mapping by introducing "star operations" (i.e., element-wise multiplication), avoiding the computational complexity brought by traditional convolution operations. While maintaining a low computational burden, StarNet can effectively enhance the feature extraction ability, especially suitable for multi-scale feature fusion, thus improving the performance of the model in complex coal flow blockage detection scenarios.

[0081] Secondly, in order to further improve the feature expression ability, the present disclosure designs a Contextual Star Fusion Module (CSFM), as Figure 5As shown. This module combines the star operation in StarBlock and further enhances the model's context awareness and attention to the central region through multi-dimensional feature fusion and the Context Anchor Attention (CAA) mechanism. This design not only enables the model to better capture the characteristics of coal flow blockage state on the basis of light weight, but also effectively reduces the computational overhead of multi-level feature fusion in traditional models.

[0082] Finally, the present disclosure proposes a Weight Shared Detection Head (WSDHead), as Figure 5 shown in the detection head (head) part. This module reduces the computational amount through the unified processing and convolution operation of the feature map, and at the same time improves the real-time detection ability of the model. The weight sharing mechanism can effectively avoid repeated calculations, making the model more efficient and meeting the requirements of real-time applications.

[0083] As Figure 5 shown, the overall framework of the model includes a feature extraction backbone network (BackBone), a neck module (Neck) for enhancing feature expression, and a lightweight shared weight detection head (WSD).

[0084] The overall framework of the model is based on StarNet as the backbone network (BackBone). StarNet is a new type of neural network architecture that achieves high-dimensional non-linear feature mapping through the star operation (Star Operation, i.e., element-wise multiplication). Without complex design or carefully selected hyperparameters, its performance exceeds that of many designs, and it is lighter than the backbone network of YOLOv8n.

[0085] The backbone (BackBone) network of StarNet contains P1, P2, P3, P4, and P5, a total of 5 feature extraction modules. There is only 1 convolutional layer in P1; there are n star blocks in each of P2, P3, P4, and P5. Different sizes of StarNet are constructed by changing the number of blocks and the number of input embedding channels. After the P5 module, the network is connected to a SPPF (Spatial Pyramid Pooling - Fast) layer, which improves the inference speed by optimizing the feature map pooling process and reducing the computational amount. Then, the feature maps output by P3, P4, and the SPPF layer are fused and concatenated in the neck ( Figure 5 The concatenation operation is represented by "C" in, that is, Concat).

[0086] To enhance the model's feature expression ability for multi-scale targets and complex scenarios, the neck structure of YOLOv8 is improved. In this disclosure, a Context Star Fusion Module (CSFM) is designed and used to replace the C2f (Faster Implementation of CSP Bottleneck with 2 convolutions) module of YOLOv8. As Figure 5 shown, in the figure, the feature map is denoted by F. The feature map F1 output by the SPPF layer is upsampled (Upsample) to obtain F2, and F2 is concatenated with the output feature map of the P4 layer. The result F3 is input into the Context Star Fusion Module (CSFM) for processing. The output feature map F4 is upsampled once and then concatenated with the output feature map of the P3 layer to obtain F6. F6 is input into the CSFM module for processing to obtain F7. F7 undergoes a convolution operation to obtain the feature map F8. F8 and F4 are concatenated to obtain F9. After F9 is processed by the CSFM module and undergoes a convolution operation, F 10 is obtained. F 10 is concatenated with the feature map output by the SPPF layer to obtain F 11 . F 11 undergoes another CSFM to obtain F 12 . F7, F 13 and F 12 are input into the Weight-Sharing Detection Head (WSD) to obtain the final detection result.

[0087] The Weight-Sharing Detection Head (WSD) simplifies the calculation process by optimizing the structure, further reducing the model's parameter quantity and enhancing the real-time detection ability.

[0088] For the complex scenario of coal flow blockage recognition, it is required that the model enhances the attention ability and extraction accuracy of important features while ensuring lightweight and real-time performance. Therefore, this paper designs a Context Star Fusion Module (CSFM). By introducing the Context Anchor Attention Mechanism (CAA) in PKINet into the StarBlock structure of StarNet, the new structure is abbreviated as the StarBlock-CAA module, and this module replaces the Bottleneck part of the C3 module in YOLOv5, enabling the model to enhance the attention to multi-scale features and improve the accuracy and real-time performance in coal flow blockage recognition while maintaining lightweight.

[0089] CAA is an attention mechanism designed to enhance the feature representation ability of neural networks and is suitable for tasks that require rich context information. It not only focuses on the global context but also strengthens the features in the central region of the image through the attention mechanism, enabling the model to pay more attention to the features related to coal flow blockage and improving the recognition accuracy of the target area. Compared with traditional large-kernel convolutions, the CAA module has fewer computational requirements and parameter counts, making the model more lightweight.

[0090] The CAA module weights the input feature map through a channel attention mechanism. As Figure 6 shown, Figure 6 is a schematic diagram of the context anchor attention mechanism module proposed according to the present disclosure. First, the input feature map is subjected to average pooling (AvgPool), and after a one-dimensional convolutional operation, is obtained. Then it is input into a horizontal convolution to obtain , and after a vertical convolution obtains . Finally, through a one-dimensional convolution and a Sigmoid activation function, the channel attention factor is obtained. The final output of the CAA module is:

[0091]

[0092] Here, is the channel attention weight obtained through the above steps, which is multiplied by the original feature map to complete the channel weighting operation.

[0093] Adding the CAA module after the element-wise multiplication step of StarBlock can further improve the model's detection performance for targets in the image without significantly increasing the computational burden.

[0094] As Figure 5 shown, the StarBlock-CAA module first processes the input feature map through depthwise convolution (DW-Conv). Subsequently, the processed feature map is input into two fully connected layers (FC) respectively, and an element-wise multiplication operation is performed on the outputs of the two branches. Then, the multiplication result is enhanced through the context anchor attention mechanism (CAA) to capture the context information in the feature map. After that, a fully connected layer (FC) is applied again to the output of the CAA for feature transformation, and spatial features are further extracted through another depthwise convolution (DW-Conv). Finally, the results of all the above operations are concatenated to form the final output of the StarBlock-CAA module.

[0095] The C3 module in YOLOv5 is a module that contains multiple convolutional layers, feature fusion, and information enhancement. It consists of multiple ConvBNSiLU combination units and 1 Bottleneck layer. Among them, ConvBNSiLU is composed of a convolutional layer (Convolution), a batch normalization layer (Batch Normalization), and a SiLU (Sigmoid Linear Unit) activation function layer, aiming to improve the efficiency and stability of feature extraction. In this study, StarBlock-CAA is used to replace the Bottleneck in the C3 module to form the CSFM lightweight architecture.

[0096] This design can significantly improve the performance of the model. CSFM uses the core element multiplication operation of StarNet to effectively map the input features into a high-dimensional non-linear feature space, enhancing the model's ability to express features. At the same time, the introduced CAA module design is relatively lightweight and can capture rich context information without significantly increasing the computational burden, further enhancing feature expression.

[0097] This disclosure designs a lightweight weight-sharing detection head (WSD). By uniformly processing the feature maps through a shared convolution module, it not only reduces the number of parameters but also effectively fuses the spatial context information of different scales. This sharing mechanism avoids the redundancy caused by independent design for each layer, improving the computational efficiency and the overall consistency of the model.

[0098] According to Figure 5 the detection head network structure shown, P3, P4, and P5 respectively represent the feature maps after neck convolution and feature fusion, representing the feature maps from a smaller scale (P3) to a larger scale (P5). Each feature map is first processed by a 1x1 ConvGN unit (including a 2D convolution Conv2d, group normalization GroupNorm, and a SiLU activation function). Then, the output results of these three branches are input into two 3x3 ConvGN units for further processing. In this way, the network can use feature maps of different scales for object detection, thereby outputting the coordinates of the detection box of the object and the category of the object within the box.

[0099] The following is the specific process of WSD.

[0100] First, extract channel features:

[0101] (1)

[0102] Among them, represents the input feature map corresponding to each scale, which are P3, P4, and P5 respectively, Denote the ConvGN combination unit (such as Figure 5 shown). Each scale feature map P3, P4, P5 will generate as the convolved feature. In the formulas here and hereafter, .

[0103] Next, through the shared convolution module, the context information of different scale feature maps is further extracted:

[0104] (2)

[0105] Among them, are two consecutive convolution operations, acting on each feature map.

[0106] The feature maps of each scale are respectively used for the prediction of bounding boxes and classes. Bounding box regression branch (each scale feature map outputs a bounding box prediction):

[0107] (3)

[0108] Among them, is the discretized maximum value (granularity of the distribution) of the bounding box regression, represents the coordinates of the predicted target box.

[0109] Class prediction branch:

[0110] (4)

[0111] Among them, is the number of classes, represents the class prediction of each scale feature map.

[0112] The prediction of each scale is adjusted by scale to obtain the final bounding box regression output:

[0113] (5)

[0114] The bounding box prediction of each scale is decoded through the DFL (Distribution Focal Loss) module. The DFL module will perform the processing of distribution regression on the bounding box regression values:

[0115] (6)

[0116] The detailed steps of the DFL module are as follows,

[0117] 1. Reshape the distribution values:

[0118] (7)

[0119] Where B is the batch size, 4 represents x, y, w, h, is the discretization granularity of the distribution, and A is the number of anchor points.

[0120] 2. Softmax, apply the softmax operation to the distribution:

[0121] (8)

[0122] 3. Weighted sum, perform a weighted sum on the discretized regression distribution:

[0123] (9)

[0124] The finally obtained decoded bounding box is:

[0125] (10)

[0126] Decode the bounding box coordinates. The decoded bounding box combines the anchor points for coordinate decoding. dist2bbox represents the distance decoding function, anchors represents the anchor points, and strides represents the scale factor of each detection layer:

[0127] (11)

[0128] Perform class prediction. The class prediction passes through the Sigmoid activation to obtain the final class confidence:

[0129] (12)

[0130] By concatenating the bounding boxes and class predictions of all scales, the final output result is obtained:

[0131] (13)

[0132] The lightweight real-time detection method for coal flow blockage state based on Star-YOLOv8 of the present invention has remarkable beneficial effects. First, by introducing the innovative Context Star Fusion Module (CSFM), the accuracy of coal flow blockage detection is significantly improved. It can accurately identify and distinguish the coal flow blockage state in a complex coal mine environment, effectively reducing the situations of false detection and missed detection. Second, the adoption of the lightweight YOLOv8 model enables the detection process to run efficiently on edge devices while ensuring high accuracy, greatly enhancing the real-time performance of detection and meeting the requirements of rapid response in coal mine production sites. At the same time, due to reducing the dependence on high-performance computing hardware, computing resources and energy consumption are effectively saved, providing a more economical and environmentally friendly solution for coal mine production. The automated detection function of the system simplifies the operation process, reduces manual intervention, lowers the operation difficulty, and also improves work efficiency. In addition, the present invention can timely and accurately detect coal flow blockage problems, thus providing real-time warnings for miners, avoiding equipment damage and production stoppage accidents, effectively ensuring coal mine production safety, reducing production interruptions and equipment maintenance costs caused by blockages, and enhancing the overall production efficiency.

[0133] Those skilled in the art will readily conceive of other embodiments of the present disclosure after considering the specification and practicing the invention disclosed herein. The present disclosure is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include known common knowledge or conventional technical means in the technical field not disclosed herein. The specification and examples are only regarded as exemplary, and the true scope and spirit of the present disclosure are pointed out by the following claims.

[0134] It should be understood that the present disclosure is not limited to the exact structures described above and shown in the drawings, and various modifications and changes can be made without departing from its scope. The scope of the present disclosure is only limited by the appended claims.

[0135] It should be noted that in the description of the present disclosure, the terms "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance. In addition, in the description of the present disclosure, unless otherwise specified, the meaning of "a plurality of" is two or more.

[0136] Any process or method description in the flowchart or described in other ways herein can be understood as representing a module, segment, or part of code including one or more executable instructions for implementing a specific logical function or process. The scope of the preferred embodiments of the present disclosure includes additional implementations, where the functions can be executed in a substantially simultaneous manner or in an order opposite to that shown or discussed according to the functions involved, which should be understood by those skilled in the technical field of the embodiments of the present disclosure.

[0137] It should be understood that various parts of the present disclosure can be implemented by hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, any one or a combination of the following techniques well known in the art can be used: discrete logic circuits having logic gate circuits for implementing logical functions on data signals, application specific integrated circuits having appropriate combinational logic gate circuits, programmable gate arrays (PGAs), field programmable gate arrays (FPGAs), etc.

[0138] Those of ordinary skill in the art of this technology can understand that all or part of the steps carried by the method of the above embodiments can be completed by instructing relevant hardware through a program, and the program can be stored in a computer-readable storage medium. When the program is executed, it includes one or a combination of the steps of the method embodiments.

[0139] In addition, in each embodiment of the present disclosure, each functional unit can be integrated in a processing module, or each unit can exist physically alone, or two or more units can be integrated in a module. The above integrated module can be implemented in the form of hardware or in the form of a software functional module. When the above integrated module is implemented in the form of a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0140] The above-mentioned storage medium can be a read-only memory, a magnetic disk, an optical disc, etc.

[0141] In the description of this specification, the description with reference to terms such as "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present disclosure. In this specification, the schematic representations of the above terms are not necessarily referring to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in a suitable manner in any one or more embodiments or examples.

[0142] Although the embodiments of the present disclosure have been shown and described above, it can be understood that the above embodiments are exemplary and should not be construed as limiting the present disclosure. Those of ordinary skill in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present disclosure.

Claims

1. A method for identifying the blocked state of the coal flow of a scraper conveyor in a fully mechanized coal mining face, characterized in that, Including: Construct an initial coal flow blockage state recognition model, where the initial coal flow blockage state recognition model is constructed based on the YOLOv8n object detection algorithm, the StarNet backbone network, the context star fusion module, and the weight-sharing detection head. The StarNet backbone network serves as the backbone of the YOLOv8n object detection algorithm, and the context star fusion module replaces the C2f module of YOLOv8. Among them, the context star fusion module is obtained by introducing the context anchor point attention mechanism in PKINet into the StarBlock structure of StarNet, and the new structure is the StarBlock-CAA module, and the Bottleneck part of the C3 module in YOLOv5 is replaced by the StarBlock-CAA module; Obtain a target data set, where the target data set includes multiple sample coal flow images, and the sample coal flow images contain scraper conveyor coal flow blockage information, including: collecting the coal transportation video of the scraper conveyor using a working face camera, and obtaining the first data set after frame extraction, and merging the first data set and the public data to obtain the target data set; Train the initial coal flow blockage state recognition model based on the target data set to obtain a target coal flow blockage state recognition model, including: dividing the target data set into a training set, a validation set, and a test set according to a ratio, and inputting the training set, the validation set, and the test set into the initial coal flow blockage state recognition model for training and saving the model weight file with the highest accuracy; Identify the coal flow blockage state of the target scraper conveyor based on the target coal flow blockage state recognition model.

2. The method according to claim 1, characterized in that, The identifying the coal flow blockage state of the target scraper conveyor based on the target coal flow blockage state recognition model includes: Obtain the real-time coal flow image of the target scraper conveyor based on a preset time interval; Perform recognition processing on the real-time coal flow image based on the target coal flow blockage state recognition model to obtain a coal flow blockage state recognition result.

3. The method according to claim 2, characterized in that, The method further includes: When the coal flow blockage state recognition result indicates that there is a coal flow blockage state in the real-time coal flow image, generate an alarm message based on the real-time coal flow image, and perform a coal flow blockage early warning based on the alarm message.

4. The method according to claim 3, wherein The generating an alarm message based on the real-time coal flow image includes: Determine the coal pile volume of the target scraper conveyor in a blocked state according to the real-time coal flow image, and generate the alarm message according to the coal pile volume.

5. The method according to claim 3, characterized in that, The generating an alarm message based on the real-time coal flow image includes: Determine the coal pile position of the target scraper conveyor in a blocked state according to the real-time coal flow image, and generate the alarm message according to the coal pile position.

6. The method according to claim 3, wherein The performing a coal flow blockage early warning based on the alarm message includes at least one of the following: Performing a light early warning based on the alarm message; Performing a voice early warning based on the alarm message; Generating a coal flow blockage state chart based on the alarm message, and performing a visual early warning based on the coal flow blockage state chart.

7. The method according to claim 2, wherein The method further includes: When the recognition result of the coal flow blockage state indicates that there is a blocked coal flow in the real-time coal flow image, a preset control instruction is sent to the target scraper conveyor, where the preset control instruction is used to instruct the target scraper conveyor to stop running.

Images