A coal mine fire monitoring and analysis method based on image recognition

Abstract

The present invention belongs to the technical field of underground coal mine fire monitoring, and specifically is a coal mine underground fire monitoring and analysis method based on image recognition. The method adopts a guide rail mobile deployment form for the monitoring equipment in the target coal mine underground area, and plans the deployment strategy of the monitoring equipment in the target coal mine underground area on the day based on the detection results of the operating health status of each monitoring equipment. The real-time all-round images collected by each available monitoring equipment for its covered monitoring sub-area are used to perceive whether there is a potential fire risk manifestation, and the fire risk concrete area block corresponding to each risk sub-area is locked and the corresponding available monitoring equipment is moved to capture the detailed image. Through dynamic texture tracking processing and multi-spectral recognition and detection processing, the fire occurrence probability of the fire risk concrete area block corresponding to each risk sub-area is evaluated, and the fire alarm sub-area is identified to trigger the target coal mine underground area fire warning, so as to realize the accurate evaluation and timely warning of the coal mine underground fire risk.

Description

Technical Field

[0001] The invention belongs to the technical field of underground coal mine fire monitoring, and specifically relates to an underground coal mine fire monitoring and analysis method based on image recognition. Background Art

[0002] As an important energy resource, coal plays a vital role in the development of the global economy. However, the working environment in coal mines is extremely complex and dangerous. There are not only a large number of flammable materials, such as coal, wood, oil, etc., but also flammable and explosive gases such as gas. In view of this, the monitoring and control of fire accidents has always been the core of safety management in coal mines.

[0003] With the continuous progress and development of science and technology, image recognition technology has gradually penetrated into the field of underground fire monitoring and management in coal mines. At present, some solutions related to underground fire monitoring in coal mines involving image recognition have also appeared in the prior art. For example, the underground safety detection method and system of coal mines based on three-dimensional face image recognition, which is published in China with the patent number CN118552911A, can identify and locate the open fire in the image based on the YOLO model, judge the spread direction and speed of the fire according to the motion trajectory, obtain fire monitoring data, extract features and perform morphological analysis on the geological structure in the collected geological data, obtain geological monitoring data, associate and integrate the fire monitoring data and the geological monitoring data through timestamp and spatial position information to obtain the first associated information and the second associated information, determine the real-time position and behavior status of each miner to judge whether the miner is in a dangerous area, and formulate and optimize safety decisions in coal mines. It not only realizes the accuracy and real-time performance of underground fire monitoring in coal mines under image recognition technology, but also can respond quickly to fire emergencies to ensure the safety of miners.

[0004] However, the above scheme still has limitations in the monitoring and identification of underground coal mine fires. Specifically, this scheme mainly uses infrared cameras and smoke sensors as auxiliary equipment for underground coal mine fire monitoring equipment to capture the color, shape and smoke distribution pattern of the flame under the open fire state to identify the position of the open fire in the positioning image.

[0005] On the one hand, because infrared cameras mainly monitor the entire underground coal mine area, the clarity and pertinence of the images they capture are relatively poor, so they only detect fires when they develop to a certain extent, and ignore the capture of abnormal performance in the early stages of fires. For example, when a fire just breaks out, the brightness and temperature of the fire area increase relative to the surrounding areas.

[0006] On the other hand, the images captured by the overall coal mine underground area monitoring are not detailed enough to conduct a sufficient and in-depth analysis of the location of the fire, and thus cannot accurately understand the authenticity of the fire in the coal mine.

[0007] In addition, this solution uses smoke sensors as auxiliary equipment, and the response speed and accuracy of smoke sensors are also subject to certain limitations. In some cases, smoke may not be generated or diffused to the location of the sensor in a timely manner, resulting in missed reports. Summary of the invention

[0008] In order to overcome the shortcomings of the background technology, an embodiment of the present invention provides a coal mine underground fire monitoring and analysis method based on image recognition, which can effectively solve the problems involved in the above-mentioned background technology.

[0009] The purpose of the present invention can be achieved through the following technical solutions: A coal mine underground fire monitoring and analysis method based on image recognition, including: S1. At a preset time point on the day, all monitoring equipment deployed on the guide rail above the target coal mine underground area is uniformly recovered, and the operating health status of each monitoring equipment is detected, so as to plan the monitoring equipment deployment strategy for the target coal mine underground area on the day, including determining the monitoring positions of each available monitoring equipment and its guide rail and covering the monitoring sub-area.

[0010] S2. According to the monitoring equipment deployment strategy for the underground area of ​​the target coal mine on that day, each available monitoring equipment moves along the guide rail to its corresponding guide rail monitoring station, and continuously performs all-round image acquisition for its covered monitoring sub-area.

[0011] S3. Based on the real-time all-round images collected by each available monitoring device for the monitoring sub-area it covers, sense whether there is potential fire risk in the corresponding monitoring sub-area covered by each available monitoring device, assign a risk sub-area label to the corresponding covered sub-area of ​​each available monitoring device that currently senses potential fire risk, and lock the fire risk concrete area block corresponding to each risk sub-area.

[0012] S4. Move the available monitoring equipment corresponding to each risk sub-area to the guide rail position corresponding to its fire risk concrete area block, capture the detailed image of the fire risk concrete area block and perform dynamic texture tracking processing and multi-spectral recognition and detection processing, evaluate the fire occurrence probability of each risk sub-area corresponding to the fire risk concrete area block, and upgrade the label of each risk sub-area whose fire occurrence probability is greater than the preset probability threshold to a fire alarm sub-area.

[0013] S5. Identify the fire alarm sub-region label and use it to trigger the target coal mine underground area fire warning, and feedback the location of the fire risk concrete area block corresponding to each current fire alarm sub-region.

[0014] Compared with the prior art, the embodiments of the present invention have at least the following advantages or beneficial effects: (1) The present invention deploys the monitoring equipment in the form of a guide rail movement, so that the monitoring equipment can move flexibly along a predetermined path. It can not only capture the all-round image of the monitoring sub-area covered by the monitoring equipment and achieve full coverage of the underground area of ​​the coal mine, but also greatly facilitate the subsequent collection and in-depth analysis of detailed images of the fire risk concrete area blocks.

[0015] (2) Before the execution of the coal mine underground fire monitoring task on the same day, the present invention combines the image acquisition function test, the mobile function test, the communication function test and the power management function test to develop a comprehensive operation health status detection mechanism for the monitoring equipment, thereby ensuring the high accuracy and authenticity of the images collected by the coal mine underground fire monitoring on the same day, thereby effectively protecting the fire safety in the coal mine.

[0016] (3) The present invention captures the relatively bright source and relatively high heat source characteristics of the early stage of the fire in the form of an image to sense whether there are potential fire risk manifestations in the corresponding monitoring sub-areas covered by each available monitoring device, efficiently screen each risk sub-area and lock its corresponding fire risk concrete area block, thereby achieving early warning of fire risks.

[0017] (4) The present invention moves the available monitoring equipment to the guide rail position corresponding to the fire risk concrete area block, captures the detailed image of the fire risk concrete area block and performs dynamic texture tracking processing and multi-spectral recognition and detection processing, thereby providing more detailed and accurate data support for fire risk assessment and ensuring the reliability of the fire occurrence probability assessment results of the fire risk concrete area block corresponding to each risk sub-area. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The present invention is further described using the accompanying drawings, but the embodiments in the accompanying drawings do not constitute any limitation to the present invention. A person skilled in the art can obtain other drawings based on the following drawings without creative work.

[0019] Figure 1 The present invention is a flowchart of the steps for implementing the method. DETAILED DESCRIPTION

[0020] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0021] Reference Figure 1As shown, the present invention provides a coal mine underground fire monitoring and analysis method based on image recognition, including: S1. At a preset time point on the day, all monitoring equipment deployed on the guide rail above the target coal mine underground area are uniformly recovered, and the operating health status of each monitoring equipment is detected, so as to plan the monitoring equipment deployment strategy for the target coal mine underground area on the day, including determining the monitoring stations of each available monitoring equipment and its guide rail and covering the monitoring sub-area.

[0022] Specifically, the detection of the operating health status of each monitoring device includes: performing image acquisition function test, mobile function test, communication function test and power management function test on each monitoring device in turn, evaluating the image acquisition quality index, mechanical movement accuracy index, data transmission stability index and power supply stability index of each monitoring device, and comparing them with the standard specifications of the fire monitoring equipment licensed for use stored in the WEB cloud. The standard specifications include the corresponding compliance thresholds of image acquisition quality indicators, mechanical movement accuracy indicators, data transmission stability indicators and power supply stability indicators. If all the evaluation indicators of a monitoring device are greater than or equal to their corresponding compliance thresholds, the output of the operating health status detection result of the monitoring device is normal, otherwise the output is abnormal, so as to detect the operating health status of each monitoring device.

[0023] It should be noted that the specific process of the above-mentioned image acquisition function test for each monitoring device to evaluate the image acquisition quality index is as follows: start each monitoring device to perform image acquisition for the preset test scene under various simulated light conditions in the coal mine, and then obtain each test image of each monitoring device, import it into the image software, compare it with the preset test scene standard image pre-stored in the image software, and use the image software measurement tool to obtain the contrast of each test image of each monitoring device , clarity and color reproduction , is the number of each monitoring device, , To monitor the number of devices, is the number of each test image, , For the number of test images, set the influence weight of each test image corresponding to the type of simulated light condition ,Require , according to the formula

[0024] Get the image acquisition quality index of each monitoring device, where , They are the contrast and clarity of the standard images of the preset test scene. They are respectively a preset reasonable contrast deviation threshold and a preset reasonable clarity deviation threshold.

[0025] The specific process of the above-mentioned mobile function test for each monitoring device to evaluate the mechanical motion accuracy index is: a movement instruction is issued to each monitoring device to advance a preset distance at the current guide rail position and stay for a preset time before returning to the origin, and the deviation distance of the staying position and the deviation distance of returning to the origin during the execution of the mobile instruction of each monitoring device are observed, and the deviation distance is analyzed with the preset distance to obtain the staying position deviation ratio and the return position deviation ratio during the execution of the mobile instruction of each monitoring device, and the two are added and then inverted to obtain the mechanical motion accuracy index of each monitoring device. It should be noted that a preset constant is added to the denominator of the above-mentioned inversion process to prevent the denominator from being zero.

[0026] The specific process of conducting communication function tests on each monitoring device to evaluate the data transmission stability index is as follows: each monitoring device is required to send each communication test signal to the monitoring center in turn at a preset time interval, and the cumulative number of communication test signals sent reaches a preset number, and the transmission time of each monitoring device sending each communication test signal to the monitoring center is collected. And the completeness of the data fed back by the monitoring center after receiving it , is the number of each communication test signal, , according to the formula Get the data transmission stability index of each monitoring device, Preset a reasonable transmission time threshold for the communication test signal. Test signal quantity for communication.

[0027] The specific process of the power management function test for each monitoring device to evaluate the power supply stability index is as follows: record the initial power of each monitoring device, place it in a simulated working environment underground in a coal mine and carry out simulated operation of each working mode for a fixed time. Each working mode includes standby, image acquisition and image transmission, and collect the power failure rate of each monitoring device during the simulated operation of each working mode. , voltage fluctuation value and current fluctuation value , is the number of each working mode, , The number of working modes, combined with the reasonable power-off rate threshold for each working mode provided by the monitoring equipment manufacturer , Reasonable voltage fluctuation threshold and reasonable current fluctuation threshold Compare and set the preset impact weight of each working mode corresponding to the power supply ,Require , according to the formula

[0028] Get the power supply stability index of each monitoring device, including is a natural constant.

[0029] Before the execution of the coal mine underground fire monitoring task on the same day, the embodiment of the present invention combines the image acquisition function test, the mobile function test, the communication function test and the power management function test to develop a comprehensive operation health status detection mechanism for the monitoring equipment, thereby ensuring the high accuracy and authenticity of the images collected by the coal mine underground fire monitoring on the same day, thereby effectively protecting the fire safety in the coal mine.

[0030] Specifically, the planning strategy for deploying monitoring equipment in the underground area of ​​the target coal mine on the day includes: marking each monitoring device that outputs normal operating health status detection results as each available monitoring device, extracting the preset length of the guide rail in the underground area of ​​the target coal mine stored in the WEB cloud, and performing a ratio analysis with the statistical number of available monitoring equipment, and using the ratio analysis result as the benchmark division length of the guide rail in the underground area of ​​the target coal mine, thereby equally dividing the guide rail in the underground area of ​​the target coal mine to obtain each equally divided guide rail segment, and corresponding each equally divided guide rail segment to each available monitoring device one by one, and the center position of each equally divided guide rail segment as the guide rail monitoring station of its corresponding available monitoring device, and the target coal mine underground sub-area where each equally divided guide rail segment is located as the coverage monitoring sub-area of ​​its corresponding available monitoring device.

[0031] S2. According to the monitoring equipment deployment strategy for the underground area of ​​the target coal mine on that day, each available monitoring equipment moves along the guide rail to its corresponding guide rail monitoring station, and continuously performs all-round image acquisition for its covered monitoring sub-area.

[0032] It should be noted that the monitoring equipment used in the present invention is a high-definition video monitoring equipment, which has multiple display modes, specifically covering visible light display mode, thermal imaging display mode, short-wave infrared display mode, medium-wave infrared display mode and long-wave infrared display mode. In the actual monitoring process, multiple modes operate in coordination to jointly monitor the covered monitoring sub-areas so as to be able to conduct in-depth analysis from the image data obtained in each switching display mode.

[0033] The embodiment of the present invention deploys the monitoring equipment in the form of a guide rail movement, so that the monitoring equipment can move flexibly along a predetermined path. It can not only capture the all-round image of the monitoring sub-area covered by the monitoring equipment in a targeted manner and achieve full coverage of the underground area of ​​the coal mine, but also provide great convenience for the subsequent collection and in-depth analysis of detailed images of the fire risk concrete area blocks.

[0034] S3. Based on the real-time all-round images collected by each available monitoring device for the monitoring sub-area it covers, sense whether there is potential fire risk in the corresponding monitoring sub-area covered by each available monitoring device, assign a risk sub-area label to the corresponding covered sub-area of ​​each available monitoring device that currently senses potential fire risk, and lock the fire risk concrete area block corresponding to each risk sub-area.

[0035] Specifically, the sensing of whether there is a potential fire risk in the corresponding monitoring sub-area covered by each available monitoring device includes: converting the omnidirectional image collected by each available monitoring device in the visible light display mode into a grayscale image, retrieving each highlighted source pixel whose pixel grayscale value is greater than a preset grayscale threshold in the omnidirectional image, searching and integrating each highlighted source pixel for a connected domain, screening each highlighted source pixel area block whose connected domain area is greater than a preset area, switching the visible light display mode to the thermal imaging display mode, and obtaining the heat source temperature value of each highlighted source pixel area block in the omnidirectional image collected by each available monitoring device and covering the corresponding monitoring sub-area. If the heat source temperature value of any highlighted source pixel area block in the omnidirectional image collected by a certain available monitoring device and covering the corresponding monitoring sub-area is greater than or equal to the preset heat source temperature warning threshold, then it is sensed that there is a potential fire risk in the corresponding monitoring sub-area covered by the available monitoring device; otherwise, it is sensed that there is no potential fire risk.

[0036] Specifically, the fire risk representation area block corresponding to each risk sub-area is a highlighted source pixel area block corresponding to its maximum heat source temperature value.

[0037] The embodiments of the present invention capture the relatively bright source and relatively high heat source characteristics at the early stage of a fire in the form of an image to sense whether there are potential fire risk manifestations in the corresponding monitoring sub-areas covered by each available monitoring device, efficiently screen each risk sub-area and lock its corresponding fire risk concrete area block, thereby achieving early warning of fire risks.

[0038] S4. Move the available monitoring equipment corresponding to each risk sub-area to the guide rail position corresponding to its fire risk concrete area block, capture the detailed image of the fire risk concrete area block and perform dynamic texture tracking processing and multi-spectral recognition and detection processing, evaluate the fire occurrence probability of each risk sub-area corresponding to the fire risk concrete area block, and upgrade the label of each risk sub-area whose fire occurrence probability is greater than the preset probability threshold to a fire alarm sub-area.

[0039] It should be noted that the above-mentioned detail image of the fire risk concrete area block specifically refers to the close-up image collected for the fire risk concrete area block, with the fire risk concrete area block as the main display content. The picture size is adapted to the size of the area block, so that the entire detail image fully presents the overall picture of the fire risk concrete area block.

[0040] Specifically, the evaluation of the probability of fire occurrence in the fire risk concrete area block corresponding to each risk sub-area includes: collecting detailed images in the visible light display mode at each unit time point within a preset time period of the fire risk concrete area block corresponding to each risk sub-area, continuously tracking the dynamic texture feature changes in the detailed images, and analyzing the fire trend compliance coefficient of the fire risk concrete area block corresponding to each risk sub-area. , is the number of each risk sub-area, , is the number of risk sub-areas.

[0041] The visible light display mode is switched to short-, medium- and long-wave infrared display modes respectively, and the fire risk verification coefficient of each risk sub-area corresponding to the fire risk concrete area block under the switched spectral mode is evaluated. .

[0042] By formula Assess the probability of fire occurrence in the fire risk concrete area blocks corresponding to each risk sub-area.

[0043] Specifically, the The analysis process includes: constructing the grayscale co-occurrence matrix of the detail image under the visible light display mode at each unit time point within the preset period of the fire risk concrete area block corresponding to each risk sub-area; according to the texture feature parameters fed back by the grayscale co-occurrence matrix, including energy, homogeneity, entropy and contrast, analyzing the texture fuzziness, texture complexity and texture roughness of the detail image under the visible light display mode within the preset period of the fire risk concrete area block corresponding to each risk sub-area, which are recorded as , according to the formula Analyze the fire trend compliance coefficient of each risk sub-area corresponding to the fire risk concrete area block, for , The texture fuzziness, texture complexity and texture roughness of the fire evolution process monitoring image stored in the WEB cloud correspond to the preset change trend matching index thresholds.

[0044] Specifically, the The analysis process includes: calculating the texture fuzziness of the detail image in the visible light display mode at each unit time point within the preset time period of each risk sub-area corresponding to the fire risk concrete area block , is the number of each unit time point in the preset time period, , according to the formula The texture fuzziness corresponding change trend index of the detail image in the visible light display mode of the fire risk concrete area block corresponding to each risk sub-area within the preset time period is obtained, where For the The risk sub-area corresponds to the fire risk concrete area block within the preset period of time. The texture blur of the detail image in the visible light display mode at each unit time point, It is the number of unit time points in the preset time period.

[0045] It should be noted that the texture blur of the detail image in the visible light display mode at each unit time point within the preset period of each risk sub-area corresponding to the fire risk concrete area block is , is obtained by extracting the entropy of the gray-level co-occurrence matrix feedback of the detail image in the visible light display mode at each unit time point within the preset period of each risk sub-area corresponding to the fire risk concrete area block Homogeneity and contrast , according to the formula

[0046] Calculated, where are the preset weights of homogeneity, entropy, and contrast corresponding to texture blur, respectively. It should be noted that the texture feature parameters fed back by the gray-level co-occurrence matrix are all standardized and their value range is .

[0047] For example, the above The specific values ​​of can be .

[0048] It should also be noted that the above The analysis process is the same as The analysis process principle is consistent with that of , that is, by extracting the energy feedback of the grayscale co-occurrence matrix of the detail image in the visible light display mode at each unit time point in the preset period of each risk sub-area corresponding to the fire risk concrete area block ,entropy and contrast , according to the formula

[0049] Calculate the texture complexity of the detail image in the visible light display mode at each unit time point within the preset time period of each risk sub-area corresponding to the fire risk concrete area block, are the preset weights of entropy, contrast, and energy corresponding to texture complexity, respectively. , further by the formula The texture complexity corresponding change trend index of the detail image in the visible light display mode of each risk sub-area corresponding to the fire risk concrete area block within the preset time period is obtained, where For the The risk sub-area corresponds to the fire risk concrete area block within the preset period of time. The texture complexity of the detail image in the visible light presentation mode at each unit time point.

[0050] For example, the above The specific values ​​of can be 0.5, 0.3, and 0.2 respectively.

[0051] By extracting the entropy of the gray-level co-occurrence matrix feedback of the detail image in the visible light display mode at each unit time point within the preset period of each risk sub-area corresponding to the fire risk concrete area block and contrast , according to the formula Calculate the texture roughness of the detail image in the visible light display mode at each unit time point within the preset time period of each risk sub-area corresponding to the fire risk concrete area block, are the preset weights of entropy and contrast corresponding to texture roughness, and are further expressed by the formula

[0052] The texture complexity corresponding change trend index of the detail image in the visible light display mode of each risk sub-area corresponding to the fire risk concrete area block within the preset time period is obtained, where For the The risk sub-area corresponds to the fire risk concrete area block within the preset period of time. The texture roughness of the detail image in the visible light presentation mode at each unit time point.

[0053] For example, the above The specific values ​​of can be 0.5 and 0.5 respectively.

[0054] Specifically, the The analysis process includes: according to the detailed images of each unit time point in the short, medium and long wave infrared display mode of the fire risk concrete image area block corresponding to each risk sub-area in the preset period of time, the temperature anomaly degree, smoke diffusion degree and thermal radiation intensity in the short, medium and long wave infrared display mode of the fire risk concrete image area block corresponding to each risk sub-area in the preset period of time are analyzed, and the accumulated values ​​of the temperature anomaly degree, smoke diffusion degree and thermal radiation intensity are used as the basic verification index of fire risk, thereby obtaining the basic verification index of fire in the short, medium and long wave infrared display mode of the fire risk concrete image area block corresponding to each risk sub-area in the preset period of time, which are recorded as , according to the formula

[0055] The fire risk verification coefficient of each risk sub-area corresponding to the fire risk concrete area block under the switching spectral mode is obtained, where These are the preset weights of short, medium and long wave infrared display modes for fire risk verification.

[0056] It should be noted that the preset weights of the above-mentioned short-, medium- and long-wave infrared display modes for fire risk verification depend on the sensitivity of each spectral display mode to the detection of fire risk characteristics in the underground working environment of the coal mine. The present invention sets the comparison relationship between the preset weights of the short-, medium- and long-wave infrared display modes for fire risk verification as follows: The basis is that the underground environment of coal mines is complex and is often accompanied by dust and smoke. Medium-wave infrared can penetrate smoke and dust to a certain extent, and has a good detection capability for smoke generated in the early stage of a fire. Short-wave infrared is more sensitive to temperature changes and can detect abnormal temperature increases in potential fire source areas, but its effect may be affected to a certain extent by interference from smoke and dust. Therefore, the weight of short-wave infrared is slightly smaller than that of medium-wave infrared. In addition, long-wave infrared mainly reflects the thermal radiation of objects and has a strong detection capability for high-temperature objects. However, in the early stage of a fire, the temperature may not have risen to a very high level, and the advantages of long-wave infrared cannot be immediately reflected. Therefore, the weight of long-wave infrared is slightly smaller than that of short-wave infrared.

[0057] For example, the above The specific values ​​of can be .

[0058] Specifically, the analysis of the temperature anomaly degree, smoke diffusion degree and thermal radiation intensity under the short-wave infrared display mode within a preset time period of the fire risk concrete area block corresponding to each risk sub-area includes: collecting the temperature value and thermal radiation value of each unit time point within the preset time period of each pixel point in the detailed image of the fire risk concrete area block corresponding to each risk sub-area under the short-wave infrared display mode, and analyzing the temperature rise coefficient, temperature deviation coefficient, radiation rise coefficient and radiation deviation coefficient within the preset time period in the detailed image of the fire risk concrete area block corresponding to each risk sub-area under the short-wave infrared display mode according to the benchmark temperature value and benchmark thermal radiation value corresponding to the image pixel in the non-fire state stored in the WEB cloud. The accumulated value of the temperature rise coefficient and the temperature deviation coefficient is used as the temperature anomaly degree, and the accumulated value of the radiation rise coefficient and the radiation deviation coefficient is used as the thermal radiation intensity.

[0059] It should be noted that the specific analysis process of the temperature rise coefficient and the temperature deviation coefficient within the preset time period in the block detail image of the fire risk concrete area corresponding to each risk sub-area under the above-mentioned short-wave infrared display mode is as follows: subtract the temperature value of each unit time point within the preset time period of each pixel point in the block detail image of the fire risk concrete area corresponding to each risk sub-area under the short-wave infrared display mode from the temperature value of the previous unit time point, and further perform ratio analysis on the calculated difference with the temperature value of the previous unit time point to obtain the temperature change rate of each unit time point within the preset time period of each pixel point in the block detail image of the fire risk concrete area corresponding to each risk sub-area under the short-wave infrared display mode relative to the previous unit time point, screen the maximum temperature change rate of each pixel point in the preset time period in the block detail image of the fire risk concrete area corresponding to each risk sub-area under the short-wave infrared display mode, and obtain the temperature rise coefficient within the preset time period in the block detail image of the fire risk concrete area corresponding to each risk sub-area under the short-wave infrared display mode through mean calculation.

[0060] Calculate the deviation temperature value of the temperature value of each pixel point in the preset time period in the detailed image of the fire risk figurative region block corresponding to each risk sub-region in the short-wave infrared display mode relative to the reference temperature value of the image pixel in the non-fire state under the short-wave infrared display mode, and further perform ratio analysis on the deviation temperature value and the reference temperature value of the image pixel in the non-fire state under the short-wave infrared display mode to obtain the temperature deviation rate of the temperature value of each pixel point in the preset time period in the detailed image of the fire risk figurative region block corresponding to each risk sub-region in the short-wave infrared display mode relative to the reference temperature value of the image pixel in the non-fire state under the short-wave infrared display mode, screen the maximum temperature deviation rate of each pixel point in the preset time period in the detailed image of the fire risk figurative region block corresponding to each risk sub-region in the short-wave infrared display mode, and obtain the preset temperature deviation rate of the detailed image of the fire risk figurative region block corresponding to each risk sub-region in the short-wave infrared display mode through mean calculation. The temperature deviation coefficient within the time period needs to be explained in particular. There are two value situations for the above-mentioned deviation temperature value. One is that when the temperature value of a certain pixel point in the detailed image of the fire risk concrete area block corresponding to a risk sub-area under the short-wave infrared display mode at a unit time point within the preset time period is less than or equal to the corresponding reference temperature value of the image pixel under the non-fire state under the short-wave infrared display mode, the corresponding deviation temperature value is automatically set to 0; the other is that when the temperature value of a certain pixel point in the detailed image of the fire risk concrete area block corresponding to a risk sub-area under the short-wave infrared display mode at a unit time point within the preset time period is greater than the corresponding reference temperature value of the image pixel under the non-fire state under the short-wave infrared display mode, the calculated difference between the temperature value of the pixel point in the detailed image of the fire risk concrete area block corresponding to the risk sub-area under the short-wave infrared display mode at the unit time point within the preset time period and the corresponding reference temperature value of the image pixel under the non-fire state under the short-wave infrared display mode is used as the deviation temperature value.

[0061] It should also be noted that the radiation rise coefficient analysis process within the preset time period in the detailed image of the fire risk concrete area block corresponding to each risk sub-area in the above-mentioned short-wave infrared display mode is consistent with the temperature rise coefficient analysis process, and the radiation deviation coefficient analysis process is consistent with the temperature deviation coefficient analysis process, which will not be elaborated here.

[0062] Identify and obtain the area growth ratio of the smoke boundary at each unit time point relative to the smoke boundary at the previous unit time point within the preset time period of the detailed image of the fire risk concrete area block corresponding to each risk sub-area under the shortwave infrared display mode, and obtain the smoke diffusion degree under the shortwave infrared display mode within the preset time period of the fire risk concrete area block corresponding to each risk sub-area through mean calculation.

[0063] It should be noted that the analysis method of the temperature anomaly degree, smoke diffusion degree and thermal radiation intensity in the medium and long wave infrared display mode within the preset time period of the fire risk concrete area block corresponding to each risk sub-area is consistent with the analysis method of the temperature anomaly degree, smoke diffusion degree and thermal radiation intensity in the short wave infrared display mode within the preset time period of the fire risk concrete area block corresponding to the above-mentioned risk sub-area, and will not be repeated here.

[0064] The embodiment of the present invention moves the available monitoring equipment to the guide rail position corresponding to the fire risk concrete area block, captures the detailed image of the fire risk concrete area block and performs dynamic texture tracking processing and multi-spectral recognition and detection processing, so as to provide more detailed and accurate data support for fire risk assessment, and ensure the reliability of the fire occurrence probability assessment results of the fire risk concrete area block corresponding to each risk sub-area.

[0065] S5. Identify the fire alarm sub-region label and use it to trigger the target coal mine underground area fire warning, and feedback the location of the fire risk concrete area block corresponding to each current fire alarm sub-region.

[0066] The above contents are merely examples and explanations of the structure of the present invention. The technicians in this technical field may make various modifications or additions to the specific embodiments described or replace them in a similar manner. As long as they do not deviate from the structure of the invention or exceed the scope defined by the present invention, they should all fall within the protection scope of the present invention.

Claims

1. A coal mine underground fire monitoring and analysis method based on image recognition, characterized in that: The method includes: S1. Collect all monitoring equipment deployed on the upper rails of the target coal mine underground area at a preset time point on the day, detect the operating health status of each monitoring equipment, and plan the deployment strategy of monitoring equipment in the target coal mine underground area on the day, including determining the monitoring stations of each available monitoring equipment and its rails and the covered monitoring sub-areas; S2. According to the monitoring equipment deployment strategy for the underground area of ​​the target coal mine on that day, each available monitoring equipment moves along the guide rail to its corresponding guide rail monitoring station, and continuously performs all-round image acquisition for its covered monitoring sub-area; S3. Based on the real-time omnidirectional images collected by each available monitoring device for the monitoring sub-area it covers, sense whether there is potential fire risk in the corresponding monitoring sub-area covered by each available monitoring device, assign a risk sub-area label to the corresponding covered sub-area of ​​each available monitoring device that currently senses potential fire risk, and lock the fire risk concrete area block corresponding to each risk sub-area; S4. Move the available monitoring equipment corresponding to each risk sub-area to the guide rail position corresponding to its fire risk concrete area block, capture the detailed image of the fire risk concrete area block and perform dynamic texture tracking processing and multi-spectral recognition detection processing, evaluate the fire occurrence probability of each risk sub-area corresponding to the fire risk concrete area block, and upgrade the label of each risk sub-area whose fire occurrence probability is greater than a preset probability threshold to a fire alarm sub-area; S5. Identify the fire alarm sub-region label and use it to trigger the target coal mine underground area fire warning, and feedback the location of the fire risk concrete area block corresponding to each current fire alarm sub-region.

2. The method for monitoring and analyzing underground coal mine fires based on image recognition according to claim 1, characterized in that: The detection of the operational health status of each monitoring device includes: performing image acquisition function tests, mobile function tests, communication function tests and power management function tests on each monitoring device in turn, evaluating the image acquisition quality index, mechanical movement accuracy index, data transmission stability index and power supply stability index of each monitoring device, and comparing them with the standard specifications of the fire monitoring equipment licensed for use stored in the WEB cloud. The standard specifications include the corresponding threshold values ​​of image acquisition quality index, mechanical movement accuracy index, data transmission stability index and power supply stability index. If all evaluation indicators of a monitoring device are greater than or equal to their corresponding threshold values, the output of the operational health status detection result of the monitoring device is normal, otherwise the output is abnormal, so as to detect the operational health status of each monitoring device.

3. The method for monitoring and analyzing underground coal mine fires based on image recognition according to claim 2 is characterized in that: The planning strategy for deploying monitoring equipment in the underground area of ​​the target coal mine on the day includes: marking each monitoring equipment that outputs normal operating health status detection results as each available monitoring equipment, extracting the preset length of the guide rail in the underground area of ​​the target coal mine stored in the WEB cloud, and performing ratio analysis on it with the statistical number of available monitoring equipment, and using the ratio analysis result as the benchmark segmentation length of the guide rail in the underground area of ​​the target coal mine, thereby equally dividing the guide rail in the underground area of ​​the target coal mine to obtain each equally divided guide rail segment, and corresponding each equally divided guide rail segment to each available monitoring equipment one by one, and the center position of each equally divided guide rail segment is used as the guide rail monitoring station position of its corresponding available monitoring equipment, and the target coal mine underground sub-area where each equally divided guide rail segment is located is used as the coverage monitoring sub-area of ​​its corresponding available monitoring equipment.

4. The method for monitoring and analyzing underground coal mine fires based on image recognition according to claim 1, characterized in that: The sensing of whether there is potential fire risk in the corresponding monitoring sub-area covered by each available monitoring device includes: converting the omnidirectional image collected by each available monitoring device in the visible light display mode into a grayscale image, retrieving each highlight source pixel whose pixel grayscale value is greater than a preset grayscale threshold in the omnidirectional image, searching and integrating each highlight source pixel for a connected domain, screening each highlight source pixel area block whose connected domain area is greater than a preset area, switching the visible light display mode to the thermal imaging display mode, and obtaining the heat source temperature value of each highlight source pixel area block in the omnidirectional image collected by each available monitoring device and covering the corresponding monitoring sub-area. If the heat source temperature value of any highlight source pixel area block in the omnidirectional image collected by a certain available monitoring device and covering the corresponding monitoring sub-area is greater than or equal to the preset heat source temperature warning threshold, then it is sensed that there is potential fire risk in the corresponding monitoring sub-area covered by the available monitoring device; otherwise, it is sensed that there is no potential fire risk.

5. The method for monitoring and analyzing underground coal mine fires based on image recognition according to claim 4 is characterized in that: The fire risk concrete area block corresponding to each risk sub-area is a highlighted source pixel area block corresponding to its maximum heat source temperature value.

6. The method for monitoring and analyzing underground coal mine fires based on image recognition according to claim 1, characterized in that: The method of evaluating the probability of fire occurrence of each risk sub-region corresponding to the fire risk concrete regional block includes: collecting detailed images of each risk sub-region corresponding to the fire risk concrete regional block at each unit time point within a preset period of time in the visible light display mode, continuously tracking the dynamic texture feature changes in the detailed images, and analyzing the fire trend compliance coefficient of each risk sub-region corresponding to the fire risk concrete regional block. , is the number of each risk sub-area, , is the number of risk sub-areas; The visible light display mode is switched to short-, medium- and long-wave infrared display modes respectively, and the fire risk verification coefficient of each risk sub-area corresponding to the fire risk concrete area block under the switched spectral mode is evaluated. ; By formula Evaluate the probability of fire occurrence in each risk sub-area corresponding to the fire risk concrete area block, where is a natural constant.

7. The method for monitoring and analyzing underground coal mine fires based on image recognition according to claim 6, characterized in that: Said The analysis process includes: constructing the grayscale co-occurrence matrix of the detail image under the visible light display mode at each unit time point within the preset period of the fire risk concrete area block corresponding to each risk sub-area; according to the texture feature parameters fed back by the grayscale co-occurrence matrix, including energy, entropy, homogeneity and contrast, analyzing the texture fuzziness, texture complexity and texture roughness of the detail image under the visible light display mode within the preset period of the fire risk concrete area block corresponding to each risk sub-area, which are recorded as , according to the formula Analyze the fire trend compliance coefficient of each risk sub-area corresponding to the fire risk concrete area block, for , The texture fuzziness, texture complexity and texture roughness of the fire evolution process monitoring image stored in the WEB cloud correspond to the preset change trend matching index thresholds.

8. The method for monitoring and analyzing underground coal mine fires based on image recognition according to claim 7, characterized in that: Said The analysis process includes: calculating the texture fuzziness of the detail image in the visible light display mode at each unit time point within the preset time period of each risk sub-area corresponding to the fire risk concrete area block , is the number of each unit time point in the preset time period. , according to the formula The texture fuzziness corresponding change trend index of the detail image in the visible light display mode of the fire risk concrete area block corresponding to each risk sub-area within the preset time period is obtained, where For the The risk sub-area corresponds to the fire risk concrete area block within the preset period of time. The texture blur of the detail image in the visible light display mode at each unit time point, It is the number of unit time points in the preset time period.

9. The method for monitoring and analyzing underground coal mine fires based on image recognition according to claim 7, characterized in that: Said The analysis process includes: according to the detailed images of each unit time point in the short, medium and long wave infrared display mode of the fire risk concrete image area block corresponding to each risk sub-area in the preset period of time, the temperature anomaly degree, smoke diffusion degree and thermal radiation intensity in the short, medium and long wave infrared display mode of the fire risk concrete image area block corresponding to each risk sub-area in the preset period of time are analyzed, and the accumulated values ​​of the temperature anomaly degree, smoke diffusion degree and thermal radiation intensity are used as the basic verification index of fire risk, thereby obtaining the basic verification index of fire in the short, medium and long wave infrared display mode of the fire risk concrete image area block corresponding to each risk sub-area in the preset period of time, which are recorded as , according to the formula The fire risk verification coefficient of each risk sub-area corresponding to the fire risk concrete area block under the switching spectral mode is obtained, where These are the preset weights of short, medium and long wave infrared display modes for fire risk verification.

10. The method for monitoring and analyzing underground coal mine fires based on image recognition according to claim 9, characterized in that: The analysis of the temperature anomaly degree, smoke diffusion degree and thermal radiation intensity under the short-wave infrared display mode within a preset time period of the fire risk concrete area block corresponding to each risk sub-area includes: collecting the temperature value and thermal radiation value of each pixel point within the preset time period in the detailed image of the fire risk concrete area block corresponding to each risk sub-area under the short-wave infrared display mode, and analyzing the temperature rise coefficient, temperature deviation coefficient, radiation rise coefficient and radiation deviation coefficient within the preset time period in the detailed image of the fire risk concrete area block corresponding to each risk sub-area under the short-wave infrared display mode according to the reference temperature value and reference thermal radiation value corresponding to the image pixel in the non-fire state stored in the WEB cloud, and using the accumulated value of the temperature rise coefficient and the temperature deviation coefficient as the temperature anomaly degree, and using the accumulated value of the radiation rise coefficient and the radiation deviation coefficient as the thermal radiation intensity; Identify and obtain the area growth ratio of the smoke boundary at each unit time point relative to the smoke boundary at the previous unit time point within the preset time period of the detailed image of the fire risk concrete area block corresponding to each risk sub-area under the shortwave infrared display mode, and obtain the smoke diffusion degree under the shortwave infrared display mode within the preset time period of the fire risk concrete area block corresponding to each risk sub-area through mean calculation.

Images