A comprehensive mining face abnormal early warning method, system, electronic device and medium

Abstract

The present invention discloses a fully mechanized coal mining face abnormal warning method, system, electronic device and medium, which relates to the field of fully mechanized coal mining face abnormal warning. The method includes: collecting real-time videos of the underground fully mechanized coal mining face; intercepting images from the real-time videos of the fully mechanized coal mining face to obtain fully mechanized coal mining face images; dehazing the fully mechanized coal mining face images to obtain dehazed images; constructing an improved YOLOv5s visual abnormal detection model; inputting the dehazed images into the trained improved YOLOv5s visual abnormal detection model to obtain detection results; the detection results are normal personnel state, normal cable dragging state, abnormal personnel state and abnormal cable dragging state; when the detection result is an abnormal personnel state or an abnormal cable dragging state, an alarm signal is generated. The present invention can timely detect the abnormal conditions of the fully mechanized coal mining face and give an alarm in time, improving the safety of the fully mechanized coal mining face.

Description

Technical Field

[0001] The present invention relates to the field of abnormal warning for fully mechanized coal mining faces, and particularly to an abnormal warning method, system, electronic device and medium for fully mechanized coal mining faces. Background Art

[0002] The traditional monitoring method is to install an image sensor on the inspection robot, collect images of the equipment, environment and personnel in the underground fully mechanized coal mining face and upload them to the centralized control center. The staff will conduct centralized processing on the image data and review whether there are abnormal conditions in the working face. However, there are many monitoring scenarios required for the fully mechanized coal mining face, and there are monitoring blind spots in this method. Moreover, there will be problems of omission and untimely response in manual review. Summary of the Invention

[0003] The purpose of the present invention is to provide an abnormal warning method, system, electronic device and medium for fully mechanized coal mining faces, which can timely detect abnormal conditions in the fully mechanized coal mining face and give an alarm in time, improving the safety of the fully mechanized coal mining face.

[0004] To achieve the above purpose, the present invention provides the following solutions:

[0005] An abnormal warning method for a fully mechanized coal mining face, the method includes:

[0006] Collect the real-time video of the underground fully mechanized coal mining face;

[0007] Intercept the image in the real-time video of the fully mechanized coal mining face to obtain the fully mechanized coal mining face image;

[0008] Dewar the fully mechanized coal mining face image to obtain a de-warped image;

[0009] Construct an improved YOLOv5s visual anomaly detection model;

[0010] Input the de-warped image into the trained improved YOLOv5s visual anomaly detection model to obtain a detection result; the detection result is the normal state of personnel, the normal state of the cable, the abnormal state of personnel and the abnormal state of the cable;

[0011] When the detection result is the abnormal state of personnel or the abnormal state of the cable, generate an alarm signal.

[0012] Optionally, the dewaring the fully mechanized coal mining face image to obtain a de-warped image specifically includes:

[0013] Obtain the underground air light value;

[0014] Calculate the minimum value of the RGB three channels of the fully mechanized coal mining face image to obtain a grayscale image;

[0015] Apply a window to the grayscale image to obtain a dark channel image;

[0016] According to the dark channel image, apply the dark channel prior algorithm to calculate the initial transmittance;

[0017] Use guided filtering to optimize the initial transmittance to obtain the target transmittance;

[0018] According to the target transmittance and the downhole air light value, perform restoration calculation on the dark channel image to obtain a defogged image.

[0019] Optionally, before the step of performing restoration calculation on the dark channel image according to the target transmittance and the downhole air light value to obtain a defogged image, it further includes:

[0020] Correct the target transmittance according to the grayscale image to obtain a corrected target transmittance.

[0021] Optionally, the training process of the improved YOLOv5s visual anomaly detection model specifically includes:

[0022] Obtain the historical monitoring video of the underground fully mechanized coal mining face;

[0023] Intercept the historical monitoring video of the underground fully mechanized coal mining face to obtain historical images;

[0024] Perform defogging on the historical images to obtain defogged historical images;

[0025] Label the dangerous areas and the detection targets of the fully mechanized coal mining face in the defogged historical images to obtain labeled historical images; the detection targets of the fully mechanized coal mining face include personnel and cable draglines;

[0026] According to the positional relationship between the detection targets of the fully mechanized coal mining face and the dangerous areas in the labeled historical images, determine the detection results of the historical images;

[0027] Use the defogged historical images as the input and the detection results of the historical images as the output to train the improved YOLOv5s visual anomaly detection model to obtain a trained improved YOLOv5s visual anomaly detection model.

[0028] Optionally, the step of determining the detection results of the historical images according to the positional relationship between the detection targets of the fully mechanized coal mining face and the dangerous areas in the labeled historical images specifically includes:

[0029] When the detection target of the working face is personnel and there is an overlap between the detection target of the fully mechanized coal mining face and the dangerous area, the detection result of the historical image is the abnormal state of the personnel;

[0030] When the detection target of the working face is a person and there is no overlap between the detection target of the fully mechanized coal mining face and the dangerous area, the detection result of the historical image is the normal state of the person;

[0031] When the detection target of the working face is a trailing cable and the detection target of the fully mechanized coal mining face is within the dangerous area, the detection result of the historical image is the normal state of the trailing cable;

[0032] When the detection target of the working face is a trailing cable and the detection target of the fully mechanized coal mining face is not within the dangerous area, the detection result of the historical image is the abnormal state of the trailing cable.

[0033] A fully mechanized coal mining face abnormal warning system is applied to the above-mentioned fully mechanized coal mining face abnormal warning method. The system includes:

[0034] An acquisition module for acquiring real-time videos of the underground fully mechanized coal mining face;

[0035] A clipping module for clipping images from the real-time videos of the fully mechanized coal mining face to obtain fully mechanized coal mining face images;

[0036] A defogging module for defogging the fully mechanized coal mining face images to obtain defogged images;

[0037] A construction module for constructing a visual anomaly detection model of improved YOLOv5s;

[0038] A detection module for inputting the defogged images into the trained visual anomaly detection model of improved YOLOv5s to obtain detection results; the detection results are the normal state of the person, the normal state of the trailing cable, the abnormal state of the person, and the abnormal state of the trailing cable;

[0039] An alarm module for generating an alarm signal when the detection result is the abnormal state of the person or the abnormal state of the trailing cable.

[0040] An electronic device includes a memory and a processor. The memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute the above-mentioned fully mechanized coal mining face abnormal warning method.

[0041] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, it implements the above-mentioned fully mechanized coal mining face abnormal warning method.

[0042] According to the specific embodiments provided by the present invention, the following technical effects are disclosed by the present invention:

[0043] The fully-mechanized mining face abnormal early warning method provided by the present invention includes: collecting real-time videos of the underground fully-mechanized mining face; intercepting images from the real-time videos of the fully-mechanized mining face to obtain fully-mechanized mining face images; dehazing the fully-mechanized mining face images to obtain dehazed images; constructing an improved YOLOv5s visual abnormal detection model; inputting the dehazed images into the trained improved YOLOv5s visual abnormal detection model to obtain detection results; the detection results are normal personnel status, normal cable dragging status, abnormal personnel status, and abnormal cable dragging status; when the detection result is abnormal personnel status or abnormal cable dragging status, an alarm signal is generated. The present invention identifies and alarms the abnormal situations of cable dragging out of the groove and personnel intrusion through the improved YOLOv5s, improving the safety of the fully-mechanized mining face. Description of the Drawings

[0044] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0045] Figure 1 It is a flow chart of the fully-mechanized mining face abnormal early warning method provided by the present invention;

[0046] Figure 2 It is a flow chart of image dehazing;

[0047] Figure 3 It is a flow chart of abnormal detection of personnel intrusion;

[0048] Figure 4 It is a flow chart of abnormal detection of cable dragging out of the groove;

[0049] Figure 5 It is a module diagram of the classification system based on non-invasive pressure-volume loop provided by the present invention.

[0050] Symbol Explanation:

[0051] 1 - Acquisition module, 2 - Interception module, 3 - Dehazing module, 4 - Construction module, 5 - Detection module, 6 - Alarm module. Detailed Embodiments

[0052] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0053] The object of the present invention is to provide a comprehensive mining face abnormal early warning method, system, electronic device and medium, which can timely detect the abnormal conditions of the comprehensive mining face and give an alarm in time, improving the safety of the comprehensive mining face.

[0054] To make the above objects, features and advantages of the present invention more obvious and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0055] Embodiment 1

[0056] As Figure 1 shown, the present invention provides a comprehensive mining face abnormal early warning method, and the method includes:

[0057] Step S1: Collect the real-time video of the underground comprehensive mining face.

[0058] Step S2: Intercept the images in the real-time video of the comprehensive mining face to obtain the comprehensive mining face images.

[0059] Step S3: Dehaze the comprehensive mining face images to obtain dehazed images; specifically, dehaze the historical images in real time to obtain a real-time dehazed video. Figure 2 is the flowchart of image dehazing; as Figure 2 shown, S3 specifically includes.

[0060] Step S31: Obtain the underground air light value.

[0061] Step S32: Calculate the minimum value of the RGB three channels of the comprehensive mining face images to obtain grayscale images.

[0062] Step S33: Apply a window to filter the grayscale images to obtain dark channel images.

[0063] Specifically, where J is the underground coal mine image, J dark is the dark channel of J, J c is one of the three channels of J, C is the three channels, and x is the coordinate of a point in the image; the window size Ω(x) selected by the present invention is 15×15.

[0064] Step S34: According to the dark channel images, apply the dark channel prior algorithm to calculate the initial transmittance; in practical applications, when the window Ω(x) is constant and the global air light is a fixed value, the initial transmittance can be directly calculated according to the dark channel prior.

[0065] Specifically, where A is the global air light, is the minimum value of I(x) within the window range; is the initial transmittance; ω is a constant introduced to make the restored image look natural and retain some fogginess in the image. Here, ω = 0.95 is adopted.

[0066] Step S35: Optimize the initial transmittance using guided filtering to obtain the target transmittance; In practical applications, guided filtering is used to refine the transmittance. Since the initial transmittance is calculated assuming the same conditions in the window Ω(x), using the initial transmittance for defogging will cause the "block phenomenon". Therefore, it is necessary to refine the transmittance.

[0067] Step S36: Perform restoration calculation on the dark channel image according to the target transmittance and the underground air light value to obtain the defogged image. In practical applications, it can be obtained by calculation based on the target transmittance in the coal mine and the coal mine underground air light value.

[0068] Specifically, where I(x) is the intensity value of the pixels of the foggy image in the coal mine collected by the image sensor; t(x) is the target transmittance underground; t0 is the minimum value set to avoid noise affecting the transmittance, and t0 = 0.1.

[0069] In addition, before step S36, it also includes correcting the target transmittance according to the grayscale image to obtain the corrected target transmittance. In practical applications, for an image containing a highlight area, since it does not conform to the dark channel prior theory, it is necessary to optimize the transmittance, increase the transmittance value, and obtain the corrected target transmittance.

[0070] Specifically, where β is a parameter, and the formula has been given; σ represents variance, σ = 1.2. If σ < 1.2, it represents a highlight area; η is the attenuation coefficient; t’ is the corrected target transmittance.

[0071] Step S4: Construct an improved YOLOv5s visual anomaly detection model.

[0072] In practical applications, the improved YOLOv5s visual anomaly detection model includes a drag cable out-of-slot anomaly detection model and a personnel intrusion anomaly detection model. Among them, the drag cable out-of-slot anomaly detection model is used to detect the drag cable out-of-slot anomaly according to the defogged image; the personnel intrusion anomaly detection model is used to detect the dangerous area of personnel intrusion.

[0073] Among them, in order to obtain a YOLOv5s anomaly detection model with better robustness, two improvements are made to YOLOv5s to obtain the improved YOLOv5s visual anomaly detection model. These two improvements include:

[0074] (1) Improvement of the feature extraction network. To improve its inference accuracy, the MobileNetV3_Large structure is used in the backbone network of YOLOv5. This structure mainly introduces an attention mechanism using the Bneck module to enhance the system's feature extraction ability. At the same time, the ReLU function and the H-Swish function are used at the front and back ends of this structure respectively, enabling it to complete more accurate detection while ensuring good real-time performance.

[0075] (2) Improvement of the feature fusion network. An adaptive attention module AAM and a feature enhancement module FEM are added to the original AF-FPN structure of YOLOv5s. The AAM module can reduce the number of feature channels and lower the information loss rate of YOLOv5s during the inference process, while the FEM module enhances the feature attributes and improves the detection speed.

[0076] Step S5: Input the dehazed image into the trained improved YOLOv5s visual anomaly detection model to obtain the detection results; the detection results are the normal state of personnel, the normal state of the tow cable, the abnormal state of personnel, and the abnormal state of the tow cable.

[0077] In practical applications, since the real-time video of the underground fully mechanized coal mining face is collected, the images input into the trained improved YOLOv5s visual anomaly detection model are also continuous real-time dehazed images, that is, the real-time dehazed video processed from the real-time video.

[0078] Step S6: Generate an alarm signal when the detection result is the abnormal state of personnel or the abnormal state of the tow cable.

[0079] In addition, an abnormal warning method for a fully mechanized coal mining face provided by the present invention also includes training the improved YOLOv5s visual anomaly detection model; the training process of the improved YOLOv5s visual anomaly detection model specifically includes:

[0080] Step S01: Obtain the historical monitoring video of the underground fully mechanized coal mining face.

[0081] Step S02: Intercept the historical monitoring video of the underground fully mechanized coal mining face to obtain historical images.

[0082] As a specific implementation, use ffmpeg to intercept the working segments of the corresponding dataset and store the video as a picture every second.

[0083] Since the tow cable out-of-groove anomaly detection model and the personnel intrusion anomaly detection model are in a parallel relationship, the datasets used for training the two models are different. The dataset for training the personnel intrusion model is obtained from 10,000 real miner images, and the dataset for training the tow cable out-of-groove model consists of simulated laboratory and real downhole tow cable images. (The tow cable out-of-groove images consist of both real and laboratory models because the number of real images is less than 10,000.)

[0084] Step S03: Dehaze the historical image to obtain a dehazed historical image.

[0085] Step S04: Label the dangerous areas and the detection targets of the fully mechanized coal mining face in the dehazed historical image to obtain a labeled historical image; the detection targets of the fully mechanized coal mining face include personnel and tow cables; specifically, it is made according to the format of the VOC2007 dataset, and the LabelImg tool is used to label the dataset to generate the XML configuration file required for training, and the XML format is converted into a TXT file through a script file.

[0086] Step S05: Determine the detection result of the historical image according to the positional relationship between the detection targets of the fully mechanized coal mining face and the dangerous areas in the labeled historical image;

[0087] Step S06: Use the dehazed historical image as the input and the detection result of the historical image as the output to train the improved YOLOv5s visual anomaly detection model to obtain a trained improved YOLOv5s visual anomaly detection model.

[0088] In practical applications, the dataset of the dehazed historical image and the corresponding detection result of the historical image is divided into a training set, a validation set, and a test set in a ratio of 8:1:1. Among them, the test set is used to evaluate the generalization ability of the final model, but it is not used as the basis for algorithm-related selections such as parameter tuning and feature selection.

[0089] Specifically, by setting its corresponding training parameters - the number of iterations is 300, the batch size is 32, and the initial learning rate is 0.001, the optimal visual anomaly detection model is finally obtained.

[0090] Among them, S05 specifically includes:

[0091] When the detection target of the working face is personnel and there is an overlap between the detection target of the fully mechanized coal mining face and the dangerous area, the detection result of the historical image is an abnormal state of personnel.

[0092] When the detection target of the working face is personnel and there is no overlap between the detection target of the fully mechanized coal mining face and the dangerous area, the detection result of the historical image is a normal state of personnel.

[0093] When the detection target of the working face is the drag cable and the detection target of the fully-mechanized coal mining face is within the dangerous area, the detection result of the historical image is the normal state of the drag cable.

[0094] When the detection target of the working face is the drag cable and the detection target of the fully-mechanized coal mining face is not within the dangerous area, the detection result of the historical image is the abnormal state of the drag cable.

[0095] As a specific implementation manner, the process of abnormal detection of personnel intrusion is as Figure 3 shown:

[0096] Use the sprite annotation assistant to draw a closed dangerous area with a polygon frame, detect and locate personnel through the improved YOLOv5s, and extract the center point coordinates of the detected and located rectangular frame. When the center point coordinates of the person in the surveillance video fall outside the dangerous area, the rectangular frame of the human body detection is marked green, and green normal is displayed in the upper left corner of the surveillance video, indicating that the personnel are working in the safe area; when the center point coordinates of the person in the surveillance video fall within the dangerous area, the rectangular frame of the human body detection is marked red, and abnormal is displayed in the upper left corner of the surveillance video, indicating that the personnel are working in the dangerous area, and voice reminder is carried out and the operation of the scraper conveyor is stopped.

[0097] As a specific implementation manner, the process of abnormal detection of drag cable out-of-groove is as Figure 4 shown:

[0098] Detect and locate the drag cable video through the improved YOLOv5s model trained by the drag cable data set. Use the sprite annotation assistant to draw a closed operating area of the drag cable with a polygon frame, and extract the center point coordinates of the detected and located rectangular frame. When the center point coordinates of the drag cable in the surveillance video fall within the track area, it indicates that the drag cable is operating normally, and the rectangular frame of the drag cable detection is marked green, and green normal is displayed in the upper left corner of the surveillance video; when the center point coordinates of the drag cable in the surveillance video fall outside the track area, it indicates that the drag cable is out of the groove, and the rectangular frame of the drag cable detection is marked red, and abnormal is displayed in the upper left corner of the surveillance video, and corresponding shutdown alarm processing is carried out.

[0099] In practical applications, cameras in mines are installed at fixed positions. Based on this installation position, it can be determined whether the image captured by the camera is a cable drag out of the trough or a personnel intrusion anomaly. Therefore, based on the image, the installation position of the camera can be determined, and based on the installation position, it can be determined whether the detection target of the image captured by the camera is a cable drag out of the trough or a personnel intrusion anomaly. Therefore, when actually applying this method, based on the camera that captures the video, it can be determined whether the video is a cable drag out of the trough video or a personnel intrusion anomaly video. When the video is a cable drag out of the trough video, the video is input into the trained improved YOLOv5s visual anomaly detection model for cable drag out of the trough anomaly detection in the cable drag out of the trough anomaly detection model, and the detection result of the cable drag out of the trough anomaly detection is obtained. An alarm for the cable drag out of the trough anomaly state is given according to this detection result; when the video is a personnel intrusion anomaly video, the video is input into the trained improved YOLOv5s visual anomaly detection model for personnel intrusion anomaly detection in the personnel intrusion anomaly detection model, and the detection result of the personnel intrusion anomaly detection is obtained. An alarm for the cable drag out of the trough anomaly state is given according to this detection result.

[0100] Embodiment 2

[0101] As Figure 5 shown, the present invention provides a fully mechanized coal mining face anomaly early warning system, which is applied to the fully mechanized coal mining face anomaly early warning method of Embodiment 1. The system includes:

[0102] An acquisition module 1 for acquiring real-time videos of the fully mechanized coal mining face underground.

[0103] An interception module 2 for intercepting images from the real-time videos of the fully mechanized coal mining face to obtain fully mechanized coal mining face images.

[0104] A defogging module 3 for defogging the fully mechanized coal mining face images to obtain defogged images.

[0105] A construction module 4 for constructing an improved YOLOv5s visual anomaly detection model.

[0106] A detection module 5 for inputting the defogged images into the trained improved YOLOv5s visual anomaly detection model to obtain detection results; the detection results are normal personnel state, normal cable state, abnormal personnel state, and abnormal cable state.

[0107] An alarm module 6 for generating an alarm signal when the detection result is an abnormal personnel state or an abnormal cable state.

[0108] In practical applications, the interception module 2 and the dehazing module 3 form a real-time image dehazing module. The detection module 5 and the real-time image dehazing module are deployed to an embedded platform. After the monitoring video of the fully mechanized coal mining face is input into the embedded platform, the embedded platform can automatically process, identify, and determine whether the drag cable is working in the cable trough and whether the staff is working in a safe area. If there is an abnormality, a red frame will be displayed and an alarm will be processed.

[0109] Embodiment III

[0110] An embodiment of the present invention provides an electronic device, including a memory and a processor. The memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute the abnormal warning method for the fully mechanized coal mining face in Embodiment I.

[0111] Optionally, the above-mentioned electronic device may be a server.

[0112] In addition, an embodiment of the present invention further provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the abnormal warning method for the fully mechanized coal mining face in Embodiment I.

[0113] In this specification, the various embodiments are described in a progressive manner. The key point of each embodiment is to illustrate the differences from other embodiments. The same or similar parts among the various embodiments can be referred to each other. For the system disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the description of the method part.

[0114] Specific examples are used in this article to elaborate on the principles and implementation manners of the present invention. The descriptions of the above embodiments are only used to help understand the method of the present invention and its core idea; at the same time, for those of ordinary skill in the art, based on the idea of the present invention, there will be changes in the specific implementation manners and application scopes. In summary, the content of this specification should not be construed as a limitation to the present invention.

Claims

1. A method for abnormal early warning in a fully mechanized coal mining face, characterized in that The method includes: Collecting real-time videos of the underground fully mechanized coal mining face; Intercepting images from the real-time videos of the fully mechanized coal mining face to obtain fully mechanized coal mining face images; Removing fog from the fully mechanized coal mining face images to obtain de-fogged images; Constructing an improved YOLOv5s visual anomaly detection model; Inputting the de-fogged images into the trained improved YOLOv5s visual anomaly detection model to obtain detection results; the detection results are normal personnel status, normal cable dragging status, abnormal personnel status, and abnormal cable dragging status; When the detection result is abnormal personnel status or abnormal cable dragging status, generating an alarm signal; The training process of the improved YOLOv5s visual anomaly detection model specifically includes: Obtaining historical monitoring videos of the underground fully mechanized coal mining face; Intercepting the historical monitoring videos of the underground fully mechanized coal mining face to obtain historical images; Removing fog from the historical images to obtain de-fogged historical images; Labeling the dangerous areas and detection targets of the fully mechanized coal mining face in the de-fogged historical images to obtain labeled historical images; the detection targets of the fully mechanized coal mining face include personnel and cable dragging; Determining the detection results of the historical images according to the positional relationship between the detection targets of the fully mechanized coal mining face and the dangerous areas in the labeled historical images; Using the de-fogged historical images as input and the detection results of the historical images as output to train the improved YOLOv5s visual anomaly detection model to obtain the trained improved YOLOv5s visual anomaly detection model; The determining the detection results of the historical images according to the positional relationship between the detection targets of the fully mechanized coal mining face and the dangerous areas in the labeled historical images specifically includes: When the detection target of the working face is personnel and there is an overlap between the detection target of the fully mechanized coal mining face and the dangerous area, the detection result of the historical image is abnormal personnel status; When the detection target of the working face is personnel and there is no overlap between the detection target of the fully mechanized coal mining face and the dangerous area, the detection result of the historical image is normal personnel status; When the detection target of the working face is cable dragging and the detection target of the fully mechanized coal mining face is within the dangerous area, the detection result of the historical image is normal cable dragging status; When the detection target of the working face is cable dragging and the detection target of the fully mechanized coal mining face is not within the dangerous area, the detection result of the historical image is abnormal cable dragging status.

2. The fully-mechanized mining face abnormal warning method according to claim 1, wherein The removing fog from the fully mechanized coal mining face images to obtain de-fogged images specifically includes: Obtaining the underground air light value; Calculating the minimum value of the RGB three channels of the fully mechanized coal mining face image to obtain a grayscale image; Applying a window to filter the grayscale image to obtain a dark channel image; Calculating the initial transmittance according to the dark channel image by applying the dark channel prior algorithm; Using guided filtering to optimize the initial transmittance to obtain the target transmittance; Performing restoration calculation on the dark channel image according to the target transmittance and the underground air light value to obtain a de-fogged image.

3. The fully-mechanized coal mining face abnormal early warning method according to claim 2, characterized in that, Before calculating the restoration of the dark channel image according to the target transmittance and the downhole air light value to obtain a defogged image, the method further includes: Correcting the target transmittance according to the grayscale image to obtain a corrected target transmittance.

4. An abnormal early warning system for a fully mechanized coal mining face, characterized in that, The system is used to implement a comprehensive mining face abnormal warning method according to any one of claims 1-3; the system includes: An acquisition module for acquiring real-time videos of the underground comprehensive mining face; An interception module for intercepting images in the real-time videos of the comprehensive mining face to obtain comprehensive mining face images; A defogging module for defogging the comprehensive mining face images to obtain defogged images; A construction module for constructing an improved YOLOv5s visual abnormal detection model; A detection module for inputting the defogged image into the trained improved YOLOv5s visual abnormal detection model to obtain a detection result; the detection result is the normal state of personnel, the normal state of the cable, the abnormal state of personnel, and the abnormal state of the cable; An alarm module for generating an alarm signal when the detection result is the abnormal state of personnel or the abnormal state of the cable.

5. An electronic device, characterized in that, It includes a memory and a processor. The memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute the comprehensive mining face abnormal warning method according to any one of claims 1 to 3.

6. A computer-readable storage medium, characterized in that, It stores a computer program, and when the computer program is executed by the processor, it implements the comprehensive mining face abnormal warning method according to any one of claims 1 to 3.

Images