Coal mine underground video monitoring method

Abstract

The invention discloses an underground coal mine video monitoring method, which relates to the technical field of image communication, and comprises the following steps of: acquiring historical video data, extracting historical key feature data for analysis, establishing a feature baseline range of an underground coal mine normal operation state, constructing an anomaly rule base for different anomaly types, and establishing a feature baseline range of the underground coal mine normal operation state; establishing a mapping relation between historical key feature data and abnormal rules, acquiring real-time video data and audio data of the underground coal mine, extracting real-time key feature data, comparing the real-time key feature data with a feature baseline range, matching corresponding rules in an abnormal rule base based on the established mapping relation, outputting an abnormal analysis result, and if it is judged that an abnormality exists, outputting the abnormal analysis result. If yes, triggering a corresponding alarm according to the exception rule, associating an exception handling process, and realizing intelligent safety monitoring by collecting multi-period underground video extraction features, establishing baseline and differentiation judgment conditions, constructing an exception library and a triggering mechanism, comparing real-time data to judge exception and performing graded response handling.

Description

Technical Field

[0001] The present invention relates to the technical field of image communication, and in particular to a video monitoring method for coal mines underground. Background Art

[0002] In the field of coal mine safety production, the application of video monitoring technology has become an important means to ensure operation safety and improve production efficiency. Through remote monitoring of the behavior of personnel, the state of equipment, and environmental changes in coal mines underground, this technology plays a key role in the following aspects: In terms of personnel management, it can identify the illegal operation behaviors of workers through video monitoring, remind and correct them in real time, and effectively reduce the safety risks caused by human factors; in the monitoring of equipment operation, it can track the operation status of shearers, belt conveyors, ventilators, and gas sensors in real time, capture potential equipment faults in advance, and avoid production accidents caused by equipment abnormalities; at the same time, video monitoring can also be used to monitor the underground environmental conditions, such as roadway ventilation efficiency, gas concentration fluctuations, etc., providing intuitive and real-time data basis for environmental safety early warning.

[0003] However, there are still many problems to be solved in the existing technical solutions: on the one hand, the environment in coal mines underground is complex and harsh, and objective factors such as poor lighting conditions, high dust concentration, and strong electromagnetic interference seriously affect the quality of video images, resulting in the difficulty of traditional video monitoring methods to accurately extract key feature data, and problems such as incorrect equipment identification and misjudgment of personnel behaviors are common, greatly reducing the accuracy and reliability of monitoring; on the other hand, the current monitoring system lacks the ability to dynamically analyze the normal operation state in coal mines underground. Usually, fixed thresholds and rules are used for abnormal judgment, and it cannot adapt to the differential changes of the normal state in different operation periods and different production scenarios, and it is very easy to generate a large number of false alarms or missed alarms, interfering with the normal safety production order; in the abnormal handling link, most of the existing technologies can only issue simple alarms, lacking a perfect abnormal rule library and an automated disposal process association mechanism, and it is impossible to provide accurate disposal guidance for different types and levels of abnormal situations, making it difficult to achieve a rapid response and effective control of safety risks, and it is difficult to meet the requirements of intelligent and refined monitoring for coal mine safety production. Summary of the Invention

[0004] The technical problems solved by the present invention are as follows: There are still many problems to be solved in the existing technical solutions. On the one hand, the underground environment of coal mines is complex and harsh, and objective factors such as poor lighting conditions, high dust concentration, and strong electromagnetic interference seriously affect the quality of video images, resulting in difficulty in accurately extracting key feature data by traditional video monitoring methods. There are common problems such as incorrect equipment identification and misjudgment of personnel behavior, which greatly reduce the accuracy and reliability of monitoring. On the other hand, the current monitoring system lacks the ability to dynamically analyze the normal operation state of underground coal mines. Usually, fixed thresholds and rules are used for anomaly judgment, and it cannot adapt to the differential changes in the normal state during different operation periods and different production scenarios, easily generating a large number of false alarms or missed alarms, interfering with the normal production safety order. In the anomaly handling link, most of the existing technologies can only issue simple alarms, lacking a perfect anomaly rule library and an automated disposal process association mechanism, unable to provide accurate disposal guidance for different types and levels of anomalies, difficult to achieve rapid response and effective control of safety risks, and difficult to meet the requirements of intelligent and refined monitoring for coal mine safety production.

[0005] To solve the above technical problems, the present invention provides the following technical solutions, including the following steps:

[0006] Step S1: Obtain historical video data and extract historical key feature data;

[0007] Step S2: Analyze the historical key feature data, establish a characteristic baseline range for the normal operation state of underground coal mines, construct an anomaly rule library for different anomaly types, and establish a mapping relationship between historical key feature data and anomaly rules;

[0008] Step S3: Obtain real-time video data and audio data of underground coal mines, and extract real-time key feature data from the real-time video data;

[0009] Step S4: Compare the real-time key feature data with the characteristic baseline range, combine the analysis results of real-time audio data, and match the corresponding rules in the anomaly rule library based on the established mapping relationship between historical key feature data and anomaly rules, and output the anomaly analysis results;

[0010] Step S5: According to the anomaly analysis results, if it is determined that there is an anomaly, trigger the corresponding alarm according to the anomaly rules and associate the anomaly disposal process.

[0011] As a preferred solution of a video monitoring method for underground coal mines according to the present invention, wherein: the historical video data includes historical video data of the operation scenarios during the early shift, middle shift, and night shift of underground coal mines;

[0012] The historical key feature data is extracted through image recognition technology, specifically including:

[0013] Decompose the historical video data into a set of image frames according to the time series;

[0014] Adopt the fixed template matching method to identify device features, and the devices include shearers, belt conveyors, ventilators, and gas sensors;

[0015] Preset reflective marking points in the roadway, and determine the device displacement deviation by calculating the pixel coordinate offset of the reflective marking points;

[0016] Extract environmental features based on gray threshold segmentation, identify the area with the gray value of the image frame higher than the first threshold as strong light, and the area lower than the second threshold as shadow;

[0017] Capture dynamic features through frame difference method, select two adjacent image frames from the image frames for pixel difference calculation, mark the area where the gray value of the pixel points between adjacent frame images changes and the number exceeds the set threshold as a moving target, and distinguish personnel and devices through contour analysis;

[0018] The image frames include the frame images in the video image sequence of the historical video data.

[0019] As a preferred solution of the underground coal mine video monitoring method described in the present invention, wherein: analyze the historical key feature data, and establish a feature baseline range for the normal operation state of the underground coal mine, specifically including:

[0020] Establish a baseline library according to the device type, and the device types include shearers, belt conveyors, ventilators, and gas sensors;

[0021] The shearer includes a drum and a cutting motor, and determine the first baseline range based on the drum height offset and the cutting motor temperature;

[0022] The belt conveyor includes a conveyor belt and rollers, and determine the second baseline range based on the conveyor belt deviation and the roller speed fluctuation;

[0023] The ventilator includes blades, and determine the third baseline range based on the blade air pressure fluctuation and the blade vibration frequency;

[0024] For the gas sensor, determine the fourth baseline range based on the gas sensor response time and zero drift;

[0025] The feature baseline range serves as the basic data for the differential judgment conditions of the device type and operation period.

[0026] As a preferred embodiment of the underground coal mine video monitoring method of the present invention, it includes: setting differential determination conditions according to equipment types and normal operation periods, specifically including:

[0027] During the morning shift, set the determination condition for the number of moving targets in the densely populated area of personnel activities;

[0028] During the middle shift, set the determination condition for the operating noise of equipment;

[0029] During the night shift, set the determination condition for the roadway lighting brightness.

[0030] As a preferred embodiment of the underground coal mine video monitoring method of the present invention, it includes: constructing an abnormal rule library for different abnormal types, and establishing the mapping relationship between the historical key feature data and the abnormal rules, specifically including:

[0031] Construct an abnormal rule library, and establish a rule system according to the abnormal levels, including first-level abnormal rules, second-level abnormal rules and third-level abnormal rules;

[0032] When the first-level abnormal rule is triggered, automatically cut off the power supply of the area and start the emergency evacuation process;

[0033] When the second-level abnormal rule is triggered, suspend the conveying and notify the maintenance team process;

[0034] When the third-level abnormal rule is triggered, give a voice warning and record the information of the violator process;

[0035] Establish the mapping relationship between the feature data and the abnormal rules, including single-feature trigger rules and multi-feature associated trigger rules, and the multi-feature associated trigger rules determine the abnormal level based on the combination of logical AND and logical OR;

[0036] Based on the statistical analysis of the historical key feature data, construct an abnormal feature decision table, and the abnormal feature decision table includes the normal state determination conditions of equipment displacement deviation, gray value fluctuation, moving target duration, as well as the corresponding first-level abnormal, second-level abnormal and third-level abnormal determination conditions;

[0037] Divide the equipment status into normal status, early warning status, abnormal status and fault status;

[0038] Correspond the equipment status according to the equipment normal state determination conditions and equipment abnormal determination conditions, specifically including:

[0039] When the displacement deviation exceeds the normal state determination condition, enter the early warning state; when the displacement deviation exceeds the early warning state determination condition, enter the abnormal state; when the displacement deviation exceeds the abnormal state determination condition, enter the fault state;

[0040] When the gray - scale value fluctuation exceeds the normal - state determination condition, it enters the warning state; when the gray - scale value fluctuation exceeds the warning - state determination condition, it enters the abnormal state; when the gray - scale value fluctuation exceeds the abnormal - state determination condition, it enters the fault state;

[0041] When the duration of the moving target exceeds the normal - state determination condition, it enters the warning state;

[0042] Based on the normal - state determination condition of the device, the abnormal determination condition of the device, and the abnormal rules, a state transition table is established to record the corresponding relationship between the device state transition and the abnormal rules, specifically including:

[0043] When the device state transfers from the warning state to the abnormal state, it triggers an audible and visual alarm and notifies the patrol personnel;

[0044] When the device state transfers from the abnormal state to the fault state, it automatically cuts off the device power supply and starts the emergency shutdown process;

[0045] For the multi - feature association trigger rule, if two or more feature data simultaneously exceed the normal - state determination condition and reach the fault - state determination condition, the first - level abnormal rule is triggered; if only the abnormal - state determination condition is reached, the second - level abnormal rule is triggered; if a single feature data exceeds the normal - state determination condition but does not reach the warning - state determination condition, it is marked as the normal state; if it exceeds the warning - state determination condition but does not reach the abnormal - state determination condition, the third - level abnormal rule is triggered.

[0046] As a preferred solution of the coal - mine underground video monitoring method described in the present invention, wherein: the real - time video data includes device operation data, personnel activity data, and environmental status data;

[0047] The real - time key feature data in the video data is extracted through image recognition technology, specifically including:

[0048] Using a preset device shape template, template matching is performed on the image frames;

[0049] When the matching degree exceeds the threshold set according to the mean or standard deviation of the historical key feature data, the device is recognized and its position information is recorded;

[0050] The image frames are subjected to gray - scale processing, and the areas where the gray - scale value is lower than the threshold set according to the historical data of the normal roadway illumination are marked as shadows, and the areas higher than the threshold are marked as highlights for identifying abnormal illumination in the roadway;

[0051] By calculating the change in the pixel positions in two adjacent image frames, moving targets are detected;

[0052] When the area of the moving target exceeds the set threshold pixels, the personnel and the device are distinguished by the contour shape;

[0053] Deduplicate the extracted real-time key feature data and eliminate the duplicate identical real-time key feature data.

[0054] Number the real-time key feature data in chronological order, establish the correspondence between the real-time key feature data, the acquisition time, and the acquisition location, and store it in the database.

[0055] As a preferred solution of a coal mine underground video monitoring method according to the present invention, wherein: compare the real-time key feature data with the feature baseline range, and based on the established mapping relationship between historical key feature data and abnormal rules, match the corresponding rules in the abnormal rule library and output the abnormal analysis result, specifically including:

[0056] When comparing the real-time key feature data with the feature baseline range of the normal operation state, preferably adopt the differential determination conditions of the corresponding operation period.

[0057] Compare the equipment displacement deviation, the gray value fluctuation, and the moving target duration with the baseline differential determination conditions of the equipment type and the operation period.

[0058] As a preferred solution of a coal mine underground video monitoring method according to the present invention, wherein: the rule matching logic includes:

[0059] If a single feature data exceeds the normal state determination condition but does not reach the warning state determination condition, mark it as the normal state; if it exceeds the warning state determination condition but does not reach the abnormal state determination condition, trigger the third-level abnormal rule and mark it as the warning state.

[0060] If a single feature data reaches or exceeds the abnormal state determination condition, trigger the second-level abnormal rule and mark it as the abnormal state.

[0061] If two or more feature data simultaneously exceed the normal state determination condition and meet the corresponding fault state determination condition, trigger the first-level abnormal rule and mark it as the fault state.

[0062] The abnormal analysis result includes the abnormal type, the abnormal location, and the abnormal level, and annotate the abnormal occurrence timestamp in the abnormal analysis result.

[0063] The feature data includes equipment displacement deviation, gray value fluctuation, and moving target duration, and the determination conditions of the feature data are determined according to the abnormal feature decision table and the state transition rule.

[0064] The moving target duration is the cumulative duration of the moving target appearing in consecutive video frames.

[0065] As a preferred solution of a coal mine underground video monitoring method according to the present invention, wherein: according to the abnormal analysis result, if it is determined that there is an abnormality, the corresponding alarm is triggered according to the abnormal rule, and the abnormal handling process is associated, specifically including:

[0066] When the abnormal analysis result is a third-level abnormality, trigger the voice warning device in the local area of the coal mine underground, and mark the abnormal position with a yellow identifier on the ground monitoring center interface;

[0067] When the abnormal analysis result is a second-level abnormality, start the underground audible and visual alarm system, set the flashing frequency of the light to c times per second, and the intensity of the alarm sound reaches d decibels. At the same time, push the abnormal information to the handheld terminal of the patrol personnel;

[0068] When the abnormal analysis result is a first-level abnormality, the system automatically cuts off the power supply of the abnormal area, starts the underground emergency evacuation audible and visual indication system, sets the evacuation signal frequency to e times per second, sends an emergency alarm to the whole mine safety command center, and automatically associates the abnormal handling process;

[0069] The automatic association of the abnormal handling process includes:

[0070] If the abnormal type is equipment failure, automatically retrieve the corresponding equipment maintenance manual and push it to the maintenance team through the explosion-proof terminal in the coal mine underground;

[0071] If the abnormal type is personnel violation, automatically generate an electronic work order containing the violation time, location, and behavior photos, and send it to the safety management department;

[0072] If the abnormal type is environmental risk, immediately close the front and rear air doors of the abnormal area and switch the ventilation system to the emergency mode.

[0073] As a preferred solution of a coal mine underground video monitoring method according to the present invention, wherein: after starting the abnormal handling process, the position of the handling personnel is tracked in real time through the RFID positioning system; if the non-response time exceeds the set duration, automatically send a reminder to their direct superior;

[0074] The RFID positioning system adopts UHF band technology to achieve precise positioning of personnel positions through positioning base stations deployed at intervals in the underground roadway;

[0075] After the handling is completed, confirm the operation by scanning the equipment QR code, and the system records the handling completion time and archives the abnormal handling record.

[0076] Advantages of the present invention: By collecting multi-period historical video data and image recognition technology, accurate feature baselines and differential thresholds are established. Combining a multi-level anomaly rule library and a triggering mechanism, it effectively solves the problems of inaccurate feature extraction and high false alarm and missed alarm rates in traditional monitoring, accurately monitors equipment, environment, and personnel anomalies, and significantly reduces accident risks. At the same time, a three-level response system is constructed, corresponding alarms are triggered for different levels of anomalies and associated disposal processes are linked to achieve reasonable allocation of resources, shorten the average handling time of faults, and greatly improve the fault handling efficiency. In addition, the full-process automation reduces labor costs, and the automatically recorded anomaly handling data provides support for management optimization. Combining RFID positioning to track disposal personnel and scanning codes to file handling records facilitates reviewing and optimizing the anomaly handling process and rule thresholds, ultimately realizing intelligent, efficient, and safe production in coal mines, reducing the comprehensive management cost in the long term, and comprehensively improving the overall management level. BRIEF DESCRIPTION OF THE DRAWINGS

[0077] Figure 1 FIG. is a schematic diagram of the basic process of a coal mine underground video monitoring method provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0078] In order to make the above objects, features, and advantages of the present invention more obvious and understandable, the following detailed description of the specific embodiments of the present invention will be made in conjunction with the accompanying drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments.

[0079] Example 1, referring to Figure 1 , which is an embodiment of the present invention, provides a coal mine underground video monitoring method, including the following steps:

[0080] Step S1, obtain historical video data and extract historical key feature data;

[0081] Step S2, analyze the historical key feature data, establish the feature baseline range of the normal operation state in the coal mine underground, construct an anomaly rule library for different anomaly types, and establish the mapping relationship between the historical key feature data and the anomaly rules;

[0082] Step S3, obtain the real-time video data and audio data in the coal mine underground, and extract the real-time key feature data in the real-time video data;

[0083] Step S4, compare the real-time key feature data with the feature baseline range, combine the analysis result of the real-time audio data, and match the corresponding rules in the anomaly rule library based on the established mapping relationship between the historical key feature data and the anomaly rules, and output the anomaly analysis result;

[0084] Step S5. According to the abnormal analysis result, if it is determined that there is an abnormality, trigger the corresponding alarm according to the abnormal rules and associate the abnormal handling process.

[0085] In one embodiment, historical video data of the operation scenarios during the morning shift, the middle shift, and the night shift are collected. Image recognition technology is used to extract equipment features (including equipment type and displacement deviation), environmental features (including strong light and shadow), and dynamic features (including personnel movement targets and equipment movement targets), and historical key feature data is constructed. Then, a baseline library is established according to the equipment type, and the statistical process control method is used to determine the feature baseline range of the normal operation state of each equipment. Differentiated judgment conditions are set according to different operation periods. At the same time, an abnormal rule library including first-level abnormal rules, second-level abnormal rules, and third-level abnormal rules is constructed, and the mapping relationship between feature data and abnormal rules is established to clarify the equipment state division and the corresponding rules for state transition. During real-time monitoring, real-time video and audio data underground are obtained, and image recognition technologies such as template matching and grayscale processing are used to extract real-time key feature data, and the corresponding relationship between it and the acquisition time and location is established. The real-time key feature data is compared with the feature baseline range, and combined with real-time audio analysis, rules are matched in the abnormal rule library according to the rule matching logic, and the analysis result including the abnormal type, location, level, and timestamp is output. If it is determined that there is an abnormality, corresponding alarms are triggered according to the abnormal level, including a local voice warning triggered by a third-level abnormality, and the power supply is cut off and an emergency evacuation is started for a first-level abnormality; at the same time, the abnormal handling process is automatically associated, such as pushing a maintenance manual for equipment failures, generating an electronic work order for personnel violations, switching the ventilation system for environmental risks, and using the RFID positioning system to track the handling personnel after starting the handling process. If there is no response after the timeout, the superior is reminded, and after the handling is completed, the QR code is scanned to confirm and record and file.

[0086] The historical video data includes historical video data of the operation scenarios during the morning shift, the middle shift, and the night shift in the coal mine underground;

[0087] The historical key feature data is extracted by image recognition technology, specifically including:

[0088] Decompose the historical video data into a set of image frames according to the time series;

[0089] Use the fixed template matching method to identify equipment features. The equipment includes shearer, belt conveyor, ventilator, and gas sensor;

[0090] Preset reflective marking points in the roadway, and determine the equipment displacement deviation by calculating the pixel coordinate offset of the reflective marking points;

[0091] Extract environmental features based on grayscale threshold segmentation, identify the area where the grayscale value of the image frame is higher than the first threshold as strong light, and the area lower than the second threshold as shadow;

[0092] Capture dynamic features through inter-frame difference method. For the image frame, select two adjacent image frames for pixel difference calculation. Mark the area where the grayscale value of the pixel points between the adjacent frame images changes and the number exceeds the set threshold as a moving target, and distinguish personnel and equipment through contour analysis;

[0093] The image frame includes the frame image in the video image sequence of the historical video data.

[0094] In one embodiment, when implementing the coal mine underground video monitoring method, first collect the historical video data of the operation scenarios during the early shift, the historical video data of the operation scenarios during the middle shift, and the historical video data of the operation scenarios during the night shift. The early shift is from 6:00 to 14:00, the middle shift is from 14:00 to 22:00, and the night shift is from 22:00 to 6:00. Decompose them into a set of image frames according to the time series, and use the fixed template matching method to identify the equipment features of the shearer and the belt conveyor. Calculate the equipment displacement deviation through the pixel coordinate offset of the preset reflective identification points in the roadway. In terms of environmental feature extraction, the first threshold is set to 200, which is determined by selecting the grayscale value that can cover 90% of the normal light area according to the right-side distribution of the peak of the grayscale histogram of the image frames during the normal operation period in the historical video data; the second threshold is set to 30, which is obtained by statistically analyzing the grayscale values of the historical video shadow areas and taking the mean minus 1.5 times the standard deviation. Thus, the area where the grayscale of the image frame is higher than the first threshold is determined as strong light, and the area lower than 30 is identified as shadow. When capturing dynamic features, the set threshold is 40 pixels. Based on the statistics of the pixel grayscale change amount of adjacent frames in the historical video, take the 80th percentile value, calculate the pixel difference of two adjacent frames of images, mark the area where the grayscale value changes and the number exceeds the set threshold as a moving target, and then distinguish personnel and equipment through contour analysis.

[0095] Analyze the historical key feature data and establish the characteristic baseline range of the normal operation state of the coal mine underground, specifically including:

[0096] Establish a baseline library according to the equipment type. The equipment types include shearer, belt conveyor, ventilator, and gas sensor;

[0097] The shearer includes a drum and a cutting motor. Determine the first baseline range based on the drum height offset and the cutting motor temperature;

[0098] The belt conveyor includes a conveyor belt and rollers. Determine the second baseline range based on the conveyor belt deviation and the roller speed fluctuation;

[0099] The ventilator includes blades, and a third baseline range is determined based on the air pressure fluctuation amount of the blades and the blade vibration frequency;

[0100] A gas sensor, and a fourth baseline range is determined based on the response time and zero drift of the gas sensor;

[0101] The characteristic baseline range serves as the basic data for the differential judgment conditions of equipment types and operation periods.

[0102] In one embodiment, when establishing the characteristic baseline range of the normal operation state in the coal mine underground, differential settings need to be implemented according to equipment types: for the shearer, based on the statistical analysis of historical operation data in the past 6 months, the normal range of the drum height offset is set to ±12 cm, the fluctuation threshold is determined with a 95% confidence interval, and the cutting motor temperature is set in the range of 45°C - 75°C. According to the temperature distribution during the stable continuous operation period of the equipment, the first baseline range is formed; for the belt conveyor, by analyzing the historical operation data, the allowable range of the belt deviation is set to ±8 cm. Referring to the equipment safety operation standard and the actual working conditions, the fluctuation of the idler speed is controlled within ±10% of the rated speed. Based on the speed monitoring data during daily load changes, the second baseline range is formed; the ventilator baseline library sets the normal range of the air pressure fluctuation amount to ±150 Pa according to the historical data of air pressure and blade vibration. According to the statistical analysis of the air pressure fluctuation during stable operation in different operation periods, the blade vibration frequency is between 22 Hz and 28 Hz, and the vibration frequency distribution interval during normal operation of the equipment is taken to form the third baseline range; the gas sensor baseline library sets the response time to be greater than or equal to 12 seconds according to the sensor performance test and the actual operation data, meeting the timeliness requirements of safety monitoring, and the zero drift is controlled within ±0.03% FS. Based on the statistical analysis of the zero offset test during long-term operation, the fourth baseline range is formed. Each baseline range is determined based on the statistical analysis of historical key characteristic data, combined with the equipment safety standard and the actual operation conditions, providing a quantitative basis for the differential judgment of the equipment state in subsequent operation periods.

[0103] Differential judgment conditions are set according to equipment types and normal operation periods, specifically including:

[0104] During the morning shift period, the judgment condition for the number of moving targets in the densely populated area of personnel activities is set;

[0105] During the middle shift period, the judgment condition for the equipment operation noise is set;

[0106] During the night shift period, the judgment condition for the roadway lighting brightness is set.

[0107] In one of the embodiments, when setting different determination conditions according to equipment types and normal working periods, the specific operations are as follows: During the morning shift period, based on the personnel activity patterns in historical video data, densely populated areas for personnel activities are demarcated, such as the entrance of the working face and the shift handover point. The threshold for the number of moving targets in this area is set to be greater than or equal to 15 people, determined according to the 90% confidence interval of the same period data in the morning shift in the past 3 months. If the threshold is exceeded, an alarm is triggered; During the middle shift period, which is the period of concentrated equipment operation, the equipment operation audio of the shearer and belt conveyor is collected through audio sensors, and the frequency and intensity characteristics are extracted. The measured noise is compared with the baseline noise range for the middle shift corresponding to the equipment type. For example, the normal operating noise range of the shearer is 80 - 100 decibels. When the measured noise exceeds the baseline by ±10%, it is determined as abnormal; During the night shift period, with the core of ensuring lighting safety, based on the historical data of roadway lighting, the average brightness of the roadway is set to be not less than 20 lux. Based on the coal mine safety regulations and the lighting standards for night operations, if the measured brightness is lower than this threshold, it is determined as lighting abnormal. The determination conditions for each period are set by combining historical data statistics and safety standards to ensure the pertinence and effectiveness of monitoring during different working periods.

[0108] For different types of anomalies, an anomaly rule library is constructed, and the mapping relationship between historical key feature data and anomaly rules is established, specifically including:

[0109] Construct an anomaly rule library, and establish a rule system according to the anomaly level, including first-level anomaly rules, second-level anomaly rules, and third-level anomaly rules;

[0110] For the first-level anomaly rule, when triggered, the power supply of the area is automatically cut off and the emergency evacuation process is started;

[0111] For the second-level anomaly rule, when triggered, the conveying is paused and the maintenance team is notified;

[0112] For the third-level anomaly rule, when triggered, a voice warning is given and the information of the violating personnel is recorded;

[0113] Establish the mapping relationship between feature data and anomaly rules, including single-feature trigger rules and multi-feature associated trigger rules. The multi-feature associated trigger rules determine the anomaly level based on the combination of logical AND and logical OR;

[0114] Based on the statistical analysis of historical key feature data, an anomaly feature decision table is constructed. The anomaly feature decision table includes the normal state determination conditions for equipment displacement deviation, gray value fluctuation, and moving target duration, as well as the corresponding first-level anomaly, second-level anomaly, and third-level anomaly determination conditions;

[0115] The equipment status is divided into normal status, warning status, abnormal status, and fault status;

[0116] According to the equipment normal state determination conditions and equipment abnormal determination conditions, the corresponding equipment status is determined, specifically including:

[0117] When the displacement deviation exceeds the normal state determination condition, it enters the warning state; when the displacement deviation exceeds the warning state determination condition, it enters the abnormal state; when the displacement deviation exceeds the abnormal state determination condition, it enters the fault state;

[0118] When the gray value fluctuation exceeds the normal state determination condition, it enters the warning state; when the gray value fluctuation exceeds the warning state determination condition, it enters the abnormal state; when the gray value fluctuation exceeds the abnormal state determination condition, it enters the fault state;

[0119] When the duration of the moving target exceeds the normal state determination condition, it enters the warning state;

[0120] Based on the device normal state determination condition, device abnormal determination condition and abnormal rules, a state transition table is established to record the corresponding relationship between device state transitions and abnormal rules, specifically including:

[0121] When the device state transfers from the warning state to the abnormal state, it triggers an audible and visual alarm and notifies the patrol personnel;

[0122] When the device state transfers from the abnormal state to the fault state, it automatically cuts off the device power supply and starts the emergency shutdown process;

[0123] For the multi-feature association trigger rule, if two or more feature data simultaneously exceed the normal state determination condition and reach the fault state determination condition, the first-level abnormal rule is triggered; if only the abnormal state determination condition is reached, the second-level abnormal rule is triggered. If a single feature data exceeds the normal state determination condition but does not reach the warning state determination condition, it is marked as the normal state; if it exceeds the warning state determination condition but does not reach the abnormal state determination condition, the third-level abnormal rule is triggered.

[0124] In one of the embodiments, when constructing the abnormal rule library and establishing the mapping relationship between feature data and abnormal rules, first construct abnormal rules including primary abnormal rules, secondary abnormal rules, and tertiary abnormal rules according to the severity of the abnormality. The primary abnormality triggers automatic power cut-off of the area and emergency evacuation, the secondary abnormality triggers suspension of transportation and notification to the maintenance team, and the tertiary abnormality triggers voice warnings and violation information recording. Based on the statistical analysis of historical key feature data in the past 6 months and with a 95% confidence interval as the benchmark, construct an abnormal feature decision table: the normal range of equipment displacement deviation is set at ±10 cm. When it exceeds this range, it enters the warning state. When it reaches ±15 cm, it is determined as abnormal and triggers the secondary abnormal rule. When it exceeds ±20 cm, it enters the failure state and triggers the primary abnormal rule; the normal range of grayscale value fluctuation is ±15 grayscale units. When it exceeds, it enters the warning state. When it exceeds ±25 grayscale units, it triggers the secondary abnormality. When it exceeds ±35 grayscale units, it triggers the primary abnormality; the normal duration of the moving target does not exceed 60 seconds. When it exceeds, it enters the warning state. When it exceeds 120 seconds and does not meet the failure standard, it triggers the tertiary abnormality. When it exceeds 180 seconds and meets other failure conditions, it triggers the primary abnormality. At the same time, establish single and multi-feature associated trigger rules, determine the abnormal level based on logical combination, and divide the equipment status into normal state, warning state, abnormal state, and failure state, and clarify the corresponding relationship of state transitions. For example, when warning to abnormal, it triggers audible and visual alarms and inspection notifications. When abnormal to failure, it starts emergency shutdown. All judgment values are set in combination with historical data, coal mine safety standards, and actual working conditions of the equipment to ensure the scientific nature and operability of abnormal handling.

[0125] The real-time video data includes equipment operation data, personnel activity data, and environmental status data;

[0126] The real-time key feature data in the video data is extracted through image recognition technology, specifically including:

[0127] Use a preset equipment shape template to perform template matching on the image frames;

[0128] When the matching degree exceeds 85%, the equipment is recognized and its position information is recorded;

[0129] Perform grayscale processing on the image frames, mark the area with a grayscale value lower than the ninety-first threshold as a shadow, and mark the area higher than the ninety-second threshold as a highlight to identify abnormal illumination in the roadway;

[0130] Detect moving targets by calculating the change in pixel positions in two adjacent image frames;

[0131] When the area of the moving target exceeds the ninety-third threshold number of pixels, distinguish between personnel and equipment by contour shape;

[0132] Perform deduplication processing on the extracted real-time key feature data to eliminate duplicate real-time key feature data that appears repeatedly;

[0133] Number the real-time key feature data in chronological order, establish the correspondence between the real-time key feature data, the acquisition time, and the acquisition location, and store it in the database.

[0134] In one of the embodiments, when extracting the key features from the real-time video data, the implementation steps and specific numerical settings are as follows: First, for the equipment operation data, use the preset equipment shape templates of the shearer and the conveyor to perform image frame matching. The matching degree threshold is set to 85%. Based on the statistical data of the historical data in the past 3 months, ensure that the correct recognition rate is above 90%. If the threshold is exceeded, it is determined that the equipment exists and the location information is recorded. For the environmental status data, after gray-scale processing of the image frame, set the gray-scale thresholds to 30 (shadow) and 200 (highlight) according to the historical data of the normal roadway lighting (taking the data of the night shift period in the past 6 months). The gray-scale value lower than 30 is marked as a shadow, and the value higher than 200 is marked as a highlight for detecting abnormal lighting. In the extraction of personnel activity data, detect moving targets by calculating the pixel changes between adjacent frames. When the area of the moving target exceeds 800 pixels², referring to the historical data of the normal activity area size of personnel and equipment, when covering more than 95% of the normal activity target area, distinguish between personnel and equipment based on the aspect ratio of the contour (the aspect ratio of the personnel contour is about 1.5 - 2.5, and the aspect ratio of the equipment contour is significantly different). Finally, deduplicate the extracted real-time key feature data, number it in chronological order, establish the correspondence with the acquisition time and location, and store it in the database to ensure the accuracy and traceability of the data.

[0135] Compare the real-time key feature data with the feature baseline range. Based on the established mapping relationship between the historical key feature data and the abnormal rules, match the corresponding rules in the abnormal rule library and output the abnormal analysis results, specifically including:

[0136] When comparing the real-time key feature data with the feature baseline range of the normal operation state, preferentially use the differential judgment conditions for the corresponding operation period;

[0137] Compare the equipment displacement deviation, gray-scale value fluctuation, and moving target duration with the baseline differential judgment conditions of the equipment type and operation period.

[0138] In one embodiment, when comparing real-time critical feature data with the baseline range of normal operation state features, the principle of "time period first" is strictly followed and combined with quantitative criteria: during the morning shift, the area with intensive personnel activities is preferentially compared, such as the number of moving targets at the entrance of the working face, and the threshold is set to be greater than or equal to 15 people, determined according to the 90% confidence interval of the morning shift data in the past 3 months. If it exceeds, an early warning is triggered. At the same time, taking the height of the shearer drum as an example, the normal range of equipment displacement deviation is ±10 cm, based on the 95% confidence interval of the historical data in the past 6 months. If it exceeds, it enters the early warning state; during the middle shift, the equipment operation noise is preferentially analyzed. Taking the shearer as an example, the normal noise baseline is 80 - 100 decibels. According to the statistics of historical operation data, if the measured value exceeds the baseline by ±10%, that is, greater than 110 decibels or less than 70 decibels, it is determined to be abnormal. The roller speed fluctuation is controlled within ±10% of the rated speed. Based on the daily load monitoring data, if it exceeds, the secondary abnormal rule is triggered; during the night shift, the roadway lighting brightness is preferentially verified, and the average brightness is set to be greater than or equal to 20 lux. According to the coal mine safety regulations and night operation standards, if the measured brightness is less than 20 lux and the gray value fluctuation exceeds ±35 gray units, and the mean value of the historical shadow area data is -1.5 times the standard deviation, it is determined to be lighting abnormal. For the core data of equipment displacement deviation, gray value fluctuation and moving target duration, differential comparison is made with the equipment type and time period baseline: for example, the normal range of the belt deviation of the belt conveyor is ±8 cm (refer to the equipment safety standard). If the measured deviation during the middle shift is greater than 15 cm and accompanied by abnormal roller speed, with a fluctuation greater than ±15%, the primary abnormal rule is triggered; the normal duration of the moving target is less than or equal to 60 seconds. According to the statistics of historical video data, if a single target during the night shift lasts longer than 180 seconds and the gray value is abnormal, it is determined to be an environmental risk in combination with the logical "AND" rule. After comparison, according to the abnormal feature decision table, for example, a displacement deviation of ±20 cm triggers the emergency stop and status transfer rule, and the analysis results including the abnormal type, location, level, such as primary abnormal and timestamp, are output to provide accurate basis for disposal.

[0139] The rule matching logic includes:

[0140] If a single feature data exceeds the normal state determination condition but does not reach the early warning state determination condition, it is marked as the normal state; if it exceeds the early warning state determination condition but does not reach the abnormal state determination condition, the tertiary abnormal rule is triggered and it is marked as the early warning state;

[0141] If a single feature data reaches or exceeds the abnormal state determination condition, the secondary abnormal rule is triggered and it is marked as the abnormal state;

[0142] If two or more feature data simultaneously exceed the normal state determination condition and meet the corresponding fault state determination condition, the primary abnormal rule is triggered and it is marked as the fault state;

[0143] The abnormal analysis results include the abnormal type, abnormal location, and abnormal level, and mark the timestamp of the abnormal occurrence in the abnormal analysis results;

[0144] The characteristic data includes the equipment displacement deviation, gray value fluctuation, and moving target duration. The determination conditions of the characteristic data are determined according to the abnormal characteristic decision table and state transition rules;

[0145] The moving target duration is the cumulative duration of the moving target appearing in consecutive video frames.

[0146] In one embodiment, when executing the rule matching logic, the following criteria are implemented: If a single characteristic data (equipment displacement deviation, gray value fluctuation, moving target duration) exceeds the normal state determination condition but does not reach the warning threshold, it is marked as the normal state; if it exceeds the warning state determination condition but does not reach the abnormal threshold, the third-level abnormal rule is triggered and it is marked as the warning state; if it reaches or exceeds the abnormal state determination condition, the second-level abnormal rule is triggered and it is marked as the abnormal state; if two or more characteristic data simultaneously exceed the normal state determination condition and meet the fault state determination condition (according to the abnormal characteristic decision table and state transition rules), the first-level abnormal rule is triggered and it is marked as the fault state. The abnormal analysis results need to include the abnormal type (related to equipment, environment, or personnel), abnormal location (accurate to the acquisition location), abnormal level (first-level abnormal, second-level abnormal, third-level abnormal), and mark the timestamp of the abnormal occurrence (accurate to seconds). Among them, the moving target duration is calculated according to its cumulative duration in consecutive video frames. The determination conditions of all characteristic data are based on the abnormal characteristic decision table and state transition rules to ensure the accurate correspondence between abnormal classification and disposal processes.

[0147] According to the abnormal analysis results, if it is determined that there is an abnormality, corresponding alarms are triggered according to the abnormal rules, and the abnormal disposal process is associated, specifically including:

[0148] When the abnormal analysis result is a third-level abnormal, trigger the local area voice warning device in the coal mine underground, and mark the abnormal location with a yellow identifier on the ground monitoring center interface;

[0149] When the abnormal analysis result is a second-level abnormal, start the underground audible and visual alarm system, set the flashing frequency of the lights to c times per second, and the intensity of the alarm sound reaches d decibels. At the same time, push the abnormal information to the handheld terminals of the patrol personnel;

[0150] When the abnormal analysis result is a first-level abnormal, the system automatically cuts off the power supply of the abnormal area, starts the underground emergency evacuation audible and visual indication system, sets the evacuation signal frequency to e times per second, sends an emergency alarm to the whole mine safety command center, and automatically associates the abnormal disposal process;

[0151] The automatic association of the abnormal disposal process includes:

[0152] If the abnormal type is equipment failure, automatically retrieve the corresponding equipment maintenance manual and push it to the maintenance team through the explosion-proof terminal in the coal mine.

[0153] If the abnormal type is personnel violation, automatically generate an electronic work order containing the violation time, location, and behavior photos and send it to the safety management department.

[0154] If the abnormal type is environmental risk, immediately close the air doors before and after the abnormal area and switch the ventilation system to the emergency mode.

[0155] In one embodiment, when executing the abnormal alarm trigger and disposal process, the following standards are implemented: when the abnormal analysis result is determined to be a level-three abnormality, immediately trigger the local area voice warning device in the coal mine and mark the abnormal location with a yellow identifier on the ground monitoring center interface; if it is a level-two abnormality, start the underground audible and visual alarm system, set the light flashing frequency to 3 times per second (c is set to 3, referring to the visual prompt standard for level-two alarms in the General Technical Requirements for Coal Mine Safety Monitoring Systems), set the alarm sound intensity to 85 decibels (d is set to 85, set according to the human response sensitivity to the alarm sound and the underground noise environment), and simultaneously push the abnormal information to the handheld terminals of the inspection personnel. In the case of a level-one abnormality, the system automatically cuts off the power supply in the abnormal area, starts the underground emergency evacuation audible and visual indication system, sets the evacuation signal frequency to 5 times per second (e is set to 5, meeting the high-identifiability requirements for emergency evacuation signals), sends an emergency alarm to the whole mine safety command center, and the abnormal disposal process is automatically associated according to the abnormal type: if it is equipment failure, the system automatically retrieves the corresponding equipment maintenance manual from the knowledge base and pushes it to the maintenance team through the explosion-proof terminal in the coal mine; if it is determined to be a personnel violation, an electronic work order including the violation time, location, and behavior photos is automatically generated and sent to the safety management department; if it is an environmental risk, immediately close the air doors before and after the abnormal area and switch the ventilation system to the emergency mode to ensure effective isolation of harmful gases in the abnormal area. All disposal processes strictly follow the requirements of the coal mine safety regulations and emergency plans to ensure the timeliness and standardization of abnormal responses.

[0156] After starting the abnormal disposal process, the position of the disposal personnel is tracked in real time through the RFID positioning system; if the non-response time exceeds the set duration, a reminder is automatically sent to their direct superior.

[0157] The RFID positioning system uses UHF band technology to achieve precise positioning of personnel positions through positioning base stations deployed at intervals in the underground roadway.

[0158] After the disposal is completed, confirm the operation by scanning the equipment QR code, and the system records the disposal completion time and archives the abnormal handling record.

[0159] In one of the embodiments, after starting the abnormal handling process, the system tracks the location of the handling personnel in real time through the RFID positioning system. The specific implementation is as follows: An RFID positioning system is built using UHF band technology, and positioning base stations are deployed at intervals in the underground roadway to achieve precise positioning of personnel locations. When a handling task is issued, the system automatically obtains the RFID tag information of the handling personnel and monitors their movement trajectories in real time. If the handling personnel do not respond for more than 15 minutes (set according to the coal mine emergency response standard and the underground passage efficiency), the system will automatically send a reminder message to their direct superiors. After the handling is completed, the handling personnel need to confirm the operation by scanning the device QR code. The system will automatically record the handling completion time, and archive the information of the abnormal type, handling process, and time nodes in the database to form a complete abnormal handling record, which is convenient for subsequent review and traceability.

[0160] Through multi-period historical video data collection and image recognition technology, the present invention establishes accurate feature baselines and differential thresholds, combines a multi-level abnormal rule library and a triggering mechanism, effectively solves the problems of inaccurate traditional monitoring feature extraction and high false alarm and missed alarm rates, accurately monitors equipment, environment, and personnel abnormalities, and significantly reduces accident risks; at the same time, a three-level response system is constructed, corresponding alarms are triggered for different levels of abnormalities and associated with handling processes, realizing reasonable allocation of resources, shortening the average handling time of faults, and greatly improving the fault handling efficiency; in addition, the full-process automation reduces labor costs, and the automatically recorded abnormal handling data provides support for management optimization. Combining RFID positioning to track handling personnel and scanning codes to archive handling records is convenient for reviewing and optimizing the abnormal handling process and rule thresholds, ultimately realizing intelligent, efficient, and safe production in coal mines, reducing the comprehensive management cost in the long term, and comprehensively improving the overall management level.

[0161] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program code. Among them, the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk, or optical disk. These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured article including an instruction device, and the instruction device implements the functions specified in one process Figure 1 one process or multiple processes and / or boxes Figure 1 or the functions specified in multiple boxes.

[0162] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered by the scope of the claims of the present invention.

Claims

1. A method for video monitoring in coal mines, characterized in that, It includes the following steps: Step S1: Obtain historical video data and extract historical key feature data; Step S2: Analyze the historical key feature data, establish a characteristic baseline range for the normal operation state in the coal mine underground, construct an abnormal rule library for different abnormal types, and establish a mapping relationship between the historical key feature data and the abnormal rules; Step S3: Obtain real-time video data and audio data in the coal mine underground, and extract real-time key feature data from the real-time video data; Step S4: Compare the real-time key feature data with the characteristic baseline range, combine the analysis results of the real-time audio data, and match the corresponding rules in the abnormal rule library based on the established mapping relationship between the historical key feature data and the abnormal rules, and output the abnormal analysis result; Step S5: According to the abnormal analysis result, if it is determined that there is an abnormality, trigger the corresponding alarm according to the abnormal rules and associate the abnormal handling process.

2. The method for video monitoring in coal mines as described in claim 1, characterized in that: The historical video data includes historical video data of the operation scenes during the early shift, middle shift, and night shift in the coal mine underground; The historical key feature data is extracted by image recognition technology, specifically including: Decompose the historical video data into a set of image frames according to the time series; Use the fixed template matching method to identify equipment features, and the equipment includes a shearer, a belt conveyor, a ventilator, and a gas sensor; Set reflective marking points in the roadway, and determine the equipment displacement deviation by calculating the pixel coordinate offset of the reflective marking points; Extract environmental features based on the gray threshold segmentation, identify the area with the gray value of the image frame higher than the first threshold as strong light, and the area lower than the second threshold as shadow; Capture dynamic features by the inter-frame difference method, calculate the pixel difference of two adjacent image frames selected from the image frames, mark the area where the gray value of the pixel points between the adjacent frame images changes and the number exceeds the set threshold as a moving target, and distinguish between personnel and equipment through contour analysis; The image frame includes the frame images in the video image sequence of the historical video data.

3. The coal mine underground video monitoring method according to claim 2, wherein: Analyze the historical key feature data and establish a characteristic baseline range for the normal operation state in the coal mine underground, specifically including: Establish a baseline library according to the equipment type, and the equipment type includes a shearer, a belt conveyor, a ventilator, and a gas sensor; The shearer includes a drum and a cutting motor, and determine the first baseline range based on the drum height offset and the cutting motor temperature; The belt conveyor includes a conveyor belt and idlers, and determine the second baseline range based on the conveyor belt deviation and the idler speed fluctuation; The ventilator includes blades, and determine the third baseline range based on the blade air pressure fluctuation and the blade vibration frequency; The gas sensor, determine the fourth baseline range based on the gas sensor response time and zero drift; The characteristic baseline range is used as the basic data for the differential judgment conditions of the equipment type and operation period.

4. The method for video monitoring in coal mines according to claim 3, characterized in that: Set differential judgment conditions according to the equipment type and normal operation period, specifically including: During the early shift, set the judgment condition for the number of moving targets in the densely populated area of personnel activities; During the middle shift, set the judgment condition for the equipment operation noise; During the night shift period, determine the judgment conditions for the roadway lighting brightness.

5. The method for video monitoring in coal mines as claimed in claim 4, wherein: For different types of anomalies, construct an anomaly rule library and establish the mapping relationship between the historical key feature data and the anomaly rules, specifically including: Construct an anomaly rule library and establish a rule system according to the anomaly level, including first-level anomaly rules, second-level anomaly rules, and third-level anomaly rules; When the first-level anomaly rule is triggered, automatically cut off the power supply of the area and start the emergency evacuation process; When the second-level anomaly rule is triggered, suspend the conveying and notify the maintenance team process; When the third-level anomaly rule is triggered, issue a voice warning and record the information of the violator process; Establish the mapping relationship between the feature data and the anomaly rules, including single-feature trigger rules and multi-feature association trigger rules. The multi-feature association trigger rules determine the anomaly level based on the combination of logical AND and logical OR; Based on the statistical analysis of the historical key feature data, construct an anomaly feature decision table. The anomaly feature decision table includes the normal state judgment conditions of equipment displacement deviation, gray value fluctuation, and moving target duration, as well as the corresponding first-level anomaly, second-level anomaly, and third-level anomaly judgment conditions; Divide the equipment status into normal state, warning state, abnormal state, and fault state; Correspond the equipment status according to the equipment normal state judgment conditions and equipment anomaly judgment conditions, specifically including: When the displacement deviation exceeds the normal state judgment condition, enter the warning state; when the displacement deviation exceeds the warning state judgment condition, enter the abnormal state; when the displacement deviation exceeds the abnormal state judgment condition, enter the fault state; When the gray value fluctuation exceeds the normal state judgment condition, enter the warning state; when the gray value fluctuation exceeds the warning state judgment condition, enter the abnormal state; when the gray value fluctuation exceeds the abnormal state judgment condition, enter the fault state; When the duration of the moving target exceeds the normal state judgment condition, enter the warning state; Based on the equipment normal state judgment conditions, equipment anomaly judgment conditions, and anomaly rules, establish a state transition table for recording the corresponding relationship between equipment state transition and anomaly rules, specifically including: When the equipment state transfers from the warning state to the abnormal state, trigger an audible and visual alarm and notify the inspection personnel; When the equipment state transfers from the abnormal state to the fault state, automatically cut off the power supply of the equipment and start the emergency shutdown process; For the multi-feature association trigger rule, if two or more feature data simultaneously exceed the normal state judgment condition and reach the fault state judgment condition, trigger the first-level anomaly rule; if only the abnormal state judgment condition is reached, trigger the second-level anomaly rule. If a single feature data exceeds the normal state judgment condition but does not reach the warning state judgment condition, it is marked as the normal state; if it exceeds the warning state judgment condition but does not reach the abnormal state judgment condition, trigger the third-level anomaly rule.

6. The method for video monitoring in coal mines as described in claim 5, characterized in that: The real-time video data includes equipment operation data, personnel activity data, and environmental status data; The real-time key feature data in the video data is extracted through image recognition technology, specifically including: Use a preset equipment shape template to perform template matching on the image frames; When the matching degree exceeds the threshold set according to the mean or standard deviation of the historical key feature data, identify the device and record its location information; Perform grayscale processing on the image frame, mark the area with a grayscale value lower than the threshold set according to the historical data of normal roadway lighting as a shadow, and mark the area higher than the threshold as a highlight to identify abnormal lighting in the roadway; Detect moving targets by calculating the change in pixel positions in two adjacent image frames; When the area of the moving target exceeds the set threshold pixels, distinguish between personnel and equipment by contour shape; Perform deduplication processing on the extracted real-time key feature data to eliminate duplicate real-time key feature data that appears repeatedly; Number the real-time key feature data in chronological order, establish the correspondence between the real-time key feature data, the acquisition time, and the acquisition location, and store them in the database.

7. The coal mine underground video monitoring method according to claim 6, wherein: Compare the real-time key feature data with the feature baseline range, match the corresponding rules in the abnormal rule library based on the established mapping relationship between the historical key feature data and the abnormal rules, and output the abnormal analysis results, specifically including: When comparing the real-time key feature data with the feature baseline range of the normal operation state, preferentially adopt the differential determination conditions for the corresponding operation period; Compare the device displacement deviation, grayscale value fluctuation, and moving target duration with the baseline differential determination conditions of the device type and operation period.

8. The method for video monitoring in coal mines according to claim 7, wherein: The rule matching logic includes: If a single feature data exceeds the normal state determination condition but does not reach the warning state determination condition, mark it as the normal state; if it exceeds the warning state determination condition but does not reach the abnormal state determination condition, trigger the third-level abnormal rule and mark it as the warning state; If a single feature data reaches or exceeds the abnormal state determination condition, trigger the second-level abnormal rule and mark it as the abnormal state; If two or more feature data simultaneously exceed the normal state determination condition and meet the corresponding fault state determination condition, trigger the first-level abnormal rule and mark it as the fault state; The abnormal analysis results include the abnormal type, abnormal location, and abnormal level, and mark the abnormal timestamp in the abnormal analysis results; The feature data includes device displacement deviation, grayscale value fluctuation, and moving target duration, and the determination conditions of the feature data are determined according to the abnormal feature decision table and the state transition rule; The moving target duration is the cumulative duration of the moving target appearing in consecutive video frames.

9. The method for video monitoring in coal mines according to claim 8, characterized in that: According to the abnormal analysis results, if it is determined that there is an abnormality, trigger the corresponding alarm according to the abnormal rules and associate the abnormal handling process, specifically including: When the abnormal analysis result is a third-level abnormality, trigger the voice warning device in the local area of the coal mine underground and mark the abnormal location with a yellow identifier on the ground monitoring center interface; When the abnormal analysis result is a second-level abnormality, start the underground audible and visual alarm system, set the flashing frequency of the light to c times per second, and the alarm sound intensity reaches d decibels. At the same time, push the abnormal information to the handheld terminal of the patrol personnel; When the abnormal analysis result is a first-level abnormality, the system automatically cuts off the power supply of the abnormal area, activates the underground emergency evacuation sound and light indication system, sets the evacuation signal frequency to e times per second, sends an emergency alarm to the whole mine safety command center, and automatically associates the abnormal handling process; The automatic association of the abnormal handling process includes: If the abnormal type is equipment failure, the corresponding equipment maintenance manual is automatically retrieved and pushed to the maintenance team through the explosion-proof terminal in the coal mine; If the abnormal type is personnel violation, an electronic work order containing the violation time, location, and behavior photos is automatically generated and sent to the safety management department; If the abnormal type is environmental risk, immediately close the front and rear air doors of the abnormal area and switch the ventilation system to the emergency mode.

10. The coal mine underground video monitoring method according to claim 9, characterized in that: After starting the abnormal handling process, the position of the handling personnel is tracked in real time through the RFID positioning system; if the non-response time exceeds the set duration, a reminder is automatically sent to their direct superior; The RFID positioning system uses UHF band technology to achieve precise positioning of personnel positions through positioning base stations deployed at intervals in the underground roadway; After the handling is completed, confirm the operation by scanning the equipment QR code, and the system records the handling completion time and archives the abnormal handling record.

Images