A multi-source perception fusion-based abnormal state monitoring method for underground skip car

Abstract

The present application relates to the technical field of coal mine underground track transportation safety monitoring, and discloses an underground skip car abnormal state monitoring method based on multi-source sensing fusion. The method comprises the following steps: collecting target information in front of the vehicle, wheel-rail relationship information, vehicle body posture information, image review information and vehicle position information; performing time synchronization, preprocessing and state extraction on the multi-source data; adaptively correcting the wheel-rail distance threshold value, the posture threshold value and the image offset threshold value according to the section where the vehicle is located; respectively obtaining wheel-rail relationship abnormal results, posture abnormal results and image offset abnormal results, and performing consistency fusion judgment within a preset time window to determine risk early warning or abnormal events; detecting foreign matters in front of the vehicle and generating an alarm; when an abnormal event or a foreign matter alarm is triggered, generating an event evidence chain containing snapshot pictures, key video clips, triggering basis, position and trajectory information; and uploading the state data and the event data to a ground platform according to a hierarchical uploading strategy. The present application can adapt to the complex working conditions of underground low illumination, heavy dust and water vapor, many curves, obvious vehicle shaking, no fixed power supply and no fixed network, improve the reliability of derailment, rollover and foreign matter event identification, reduce the false alarm rate, and enhance the event review and traceability.

Description

Technical Field

[0001] This invention relates to the field of safety monitoring technology for underground track transportation in coal mines, specifically to a method for monitoring abnormal states of underground skip cars based on multi-source sensing fusion. It is applicable to monitoring for foreign objects ahead, identifying derailment anomalies, providing early warning of rollover risks, verifying event images, associating locations, wireless transmission, and platform alarms during the operation of underground skip cars and mine track transportation vehicles. Background Technology

[0002] The underground rail transport system in coal mines is a crucial component of mine production operations. Skip cars are responsible for transporting materials, equipment, and other items in underground roadways. During operation, these vehicles face risks such as unauthorized personnel entering the road ahead, obstacles intruding, cars blocking the track, and other foreign objects obstructing the path. Furthermore, abnormal wheel-rail relationships, abnormal car posture, changes in track conditions, or operational impacts can lead to safety risks such as derailment, deviation, and overturning. Failure to detect these anomalies in a timely manner can easily result in shutdowns, equipment damage, production interruptions, or even safety accidents.

[0003] Existing anomaly monitoring solutions for underground rail transport often employ a single sensor or a single identification method. For example, proximity switches or distance sensors are used to detect the relative position between the vehicle and the rails; tilt sensors are used to detect changes in the vehicle's roll and pitch angles; and cameras or video recognition methods are used to monitor obstacles ahead or the condition of the track. While these methods can achieve basic monitoring functions under certain conditions, they still have significant shortcomings in practical underground applications.

[0004] First, underground mine tunnels present complex environments, typically characterized by low illumination, heavy dust, high moisture content, complex background textures, numerous curves, and significant vehicle vibration and impact. Relying solely on visual methods is susceptible to factors such as lighting, obstructions, vibration, and lens contamination, leading to unstable recognition results. Conversely, relying solely on proximity switches or tilt sensors is easily affected by normal vehicle vibrations, uneven tracks, and changes in vehicle attitude during curves, resulting in high false alarm rates and poor interference resistance.

[0005] Secondly, existing single-sensor solutions often struggle to distinguish between normal fluctuations and genuine anomalies. For example, when a skip-load truck is operating on a curve, it will experience a certain degree of change in its vehicle attitude; when the vehicle is subjected to a short-term impact, it may also experience transient changes in tilt angle or fluctuations in wheel-rail distance. Without cross-validation and temporal continuity constraints among multi-source information, normal operating conditions can easily be misjudged as abnormal events.

[0006] Furthermore, underground skip trucks typically lack a continuous and stable fixed power supply and network environment, making it difficult to maintain stable operation for extended periods using traditional methods of continuous high-volume video uploads or long-term high-power online analysis. Without designing a suitable data processing and upload mechanism that integrates with underground power and communication conditions, system implementation will be significantly challenging.

[0007] In addition, most existing solutions only output simple alarm results and lack a unified association of information such as the basis for anomaly triggering, captured images, key video clips, location data, the segment to which it belongs, and the running trajectory. This makes it difficult to answer questions such as whether an anomaly occurred, where it occurred, what the triggering basis was, and how it happened after an anomaly occurs, which is not conducive to post-event review, cause analysis, and management closure.

[0008] Furthermore, the operational characteristics differ across different sections. The range of vehicle attitude changes, image viewpoint shifts, and wheel-rail distance fluctuations vary between straight and curved sections. Using a uniform fixed threshold often results in frequent false alarms in some sections and insufficient sensitivity in others. Therefore, a monitoring method capable of adjusting thresholds based on operational section information is needed to improve overall adaptability and reliability.

[0009] Based on the above problems, there is an urgent need to provide a multi-source sensing fusion abnormal state monitoring scheme suitable for complex downhole working conditions, so as to achieve stable identification of foreign objects in front of the skip car, derailment anomalies and rollover risks, and form a complete abnormal event evidence chain and a data hierarchical uploading mechanism adapted to downhole working conditions. Invention Patent Content

[0010] (a) Purpose of the invention

[0011] The purpose of this invention is to overcome the problems of insufficient anti-interference ability of single sensing methods, poor adaptability of fixed thresholds, incomplete abnormal evidence chains, and difficulty in long-term stable operation of high-bandwidth continuous uploading in the existing technology. It provides a method for monitoring abnormal status of skip cars in underground mines based on multi-source sensing fusion, so as to improve the accuracy, stability and traceability of abnormal status identification of skip cars under complex underground working conditions.

[0012] Another objective of this invention is to provide an adaptive judgment mechanism based on vehicle location information, enabling the system to adjust the judgment threshold according to different operating scenarios such as straight sections and curved sections, reducing false alarms caused by normal fluctuations and improving adaptability in practical engineering applications.

[0013] Another objective of this invention is to achieve effective retention, reliable transmission, and platform verification of anomaly monitoring results under the constraints of no fixed power supply and no fixed network underground, through the generation of event evidence chains and hierarchical uploading mechanisms.

[0014] (II) Methodological and Technical Solutions

[0015] To achieve the above objectives, this invention provides a method for monitoring abnormal conditions of underground skip cars based on multi-source sensing fusion, comprising the following steps:

[0016] S1, Multi-source data acquisition

[0017] Collect information on targets ahead of the vehicle, wheel-rail relationship information, vehicle attitude information, image verification information, and vehicle position information.

[0018] Specifically, the forward target information is acquired by a forward sensing device, preferably a millimeter-wave radar camera, to detect whether there are personnel, obstacles, carriages occupying the lane, or track occupancy targets in front of the vehicle; the wheel-rail relationship information is acquired by a wheel-rail relationship detection device located at the bottom of the vehicle body, preferably a proximity switch, to reflect the relative distance change or relative position state between the vehicle body and the rail; the vehicle attitude information is acquired by an attitude detection device, preferably a tilt sensor, to reflect the changes in the vehicle's roll angle and pitch angle; the image verification information is acquired by an independent image acquisition device, preferably an independent camera, to identify the positional relationship between the track line and the vehicle body edge; and the vehicle position information is acquired by an underground positioning system, preferably connected to the mine's existing UWB positioning system, to determine the vehicle's current location, section, and historical trajectory.

[0019] S2, Data Synchronization and Preprocessing

[0020] Data collected from each channel is timestamped and synchronized based on a unified time reference to form multi-source raw data with a unified time sequence.

[0021] The multi-source raw data is preprocessed, including but not limited to:

[0022] The wheel-rail relationship information is subjected to jitter filtering, discrete state smoothing, and outlier removal;

[0023] The vehicle attitude information is filtered, biased, and its changing trend is extracted.

[0024] The image verification information is subjected to denoising, enhancement, edge extraction, track line recognition, and vehicle edge recognition.

[0025] Perform target identification, distance estimation, orientation determination, and alarm area mapping on the target information ahead;

[0026] Vehicle location information is mapped to specific sections and its trajectory is organized.

[0027] Through the above processing, multi-source state data under a unified time reference is obtained.

[0028] S3, Segment Recognition and Adaptive Threshold Correction

[0029] The vehicle's current segment type is identified based on its location information, and the wheel-rail distance threshold, attitude threshold, and image offset threshold are adaptively corrected based on the segment type.

[0030] The section types include at least straight sections and curved sections, and may further include sections with varying gradients, sections where vehicles meet, and special operation sections. Different sections correspond to different normal operating fluctuation ranges, therefore different judgment thresholds are configured.

[0031] For example, when the vehicle is in a straight section, a stricter attitude threshold and image offset threshold can be used; when the vehicle is in a curved section, a larger range of normal attitude changes and image offsets is allowed to adapt to the characteristics of curved operation and reduce the risk of false alarms.

[0032] S4. Single-channel anomaly detection

[0033] Wheel-rail relationship anomaly results are obtained based on wheel-rail relationship information, attitude anomaly results are obtained based on vehicle attitude information, and image offset anomaly results are obtained based on image verification information.

[0034] The wheel-rail relationship anomaly result is used to characterize whether there is an abnormal deviation between the vehicle body and the rail. Preferably, when the proximity switch detects values ​​exceeding the normal range for multiple consecutive sampling cycles, fails to detect rails for consecutive periods, or the difference between multiple detection points exceeds a set threshold, an abnormal wheel-rail relationship is determined to exist.

[0035] Attitude anomaly results are used to characterize whether there is continuous skew or abrupt change in vehicle roll or pitch. Preferably, an attitude anomaly is determined to exist when the roll angle or pitch angle exceeds a set threshold, or the rate of attitude change continuously exceeds a set threshold, or the attitude deviates from the normal range for a long period of time.

[0036] Image offset anomaly results are used to characterize whether the vehicle body has significantly deviated from the track line. Preferably, the offset of the vehicle body edge relative to the track boundary line is calculated by identifying the track boundary line and the vehicle body edge line; when the offset exceeds a set threshold, or when the track line is partially missing and accompanied by abnormal deflection of the vehicle body edge, an image offset anomaly is determined to exist.

[0037] S5. Multi-source consistency fusion judgment within the time window

[0038] Within a preset time window, consistency fusion judgment is performed on abnormal results of wheel-rail relationship, abnormal results of attitude, and abnormal results of image offset.

[0039] The consistency judgment includes at least a time consistency judgment and a persistence judgment. The time consistency judgment is used to determine whether multiple abnormal results occur or overlap within the same preset time window; the persistence judgment is used to determine whether the abnormality persists for a set duration or a number of consecutive frames.

[0040] Preferably, when at least two types of abnormal results simultaneously meet the persistence condition within the same preset time window, it is determined to be an abnormal event; when the abnormal event determination condition is not met, but at least one type of abnormal result meets the continuous abnormal change trend condition, it is determined to be a risk warning; when the abnormality is only a short-term fluctuation and does not meet the persistence condition, it is determined to be a normal fluctuation and does not trigger an abnormal event.

[0041] This step allows us to distinguish between normal vehicle vibrations and short-term attitude changes caused by curves from actual risks of derailment, deviation, or rollover.

[0042] S6. Forward Foreign Object Detection and Alarm Judgment

[0043] The system identifies targets ahead and generates a foreign object alarm event when at least one of the following targets—personnel, obstacles, carriages occupying the aisle, or tracks—enters the preset alarm area.

[0044] Preferably, different alarm levels can be set according to the target category, the relative distance between the target and the vehicle, the orientation information, and the relative motion state, so as to improve the relevance and practicality of the alarm.

[0045] S7, Generation of Event Evidence Chain

[0046] When an event is identified as an abnormal event or a foreign object alarm event, the event evidence chain generation process is automatically triggered. The event evidence chain generation process includes at least the following:

[0047] Capture key images at the moment when an anomaly or alarm is triggered;

[0048] Save the video clip of a preset duration before the exception is triggered;

[0049] Save the video clip of a preset duration after an exception is triggered;

[0050] Record the triggering basis, including at least one of the following: abnormal wheel-rail relationship results, abnormal attitude results, abnormal image offset results, and forward target detection results;

[0051] Get and bind the location, segment, and historical trajectory of the event;

[0052] Generate a unique event number;

[0053] This forms a chain of event evidence that includes event time, event type, triggering basis, images, videos, location, and trajectory information.

[0054] S8, Tiered Upload and Cache Re-upload

[0055] Upload the corresponding data according to the tiered upload strategy. The tiered upload strategy includes:

[0056] Under normal conditions, status data is uploaded, including location data, device status data, and simplified monitoring results;

[0057] When a risk warning is in effect, a warning summary data is uploaded. The warning summary data includes the warning level, trend change information, and necessary simplified image summaries.

[0058] When an abnormal event or foreign object alarm event is triggered, event data is uploaded. The event data includes at least captured images, key video clips, triggering criteria, event location, and trajectory information.

[0059] Preferably, when the underground wireless network is interrupted, abnormal event data and key evidence data are cached locally; when the network is restored, abnormal event data is retransmitted first to ensure that key event information is not lost.

[0060] S9, Platform Receipt and Review Display

[0061] After receiving the uploaded data, the ground platform displays abnormal alarms, image evidence, triggering basis, event location, running trajectory and historical records, and supports event query, trajectory playback, statistical analysis and manual review, thereby realizing traceable management of abnormal events.

[0062] (III) System Technical Solution

[0063] This invention also provides a method for monitoring abnormal states of underground skip cars based on multi-source sensing fusion, comprising:

[0064] 1. Forward Perception Module

[0065] The system is used to collect information about targets in front of the vehicle, preferably including a millimeter-wave radar camera. The forward sensing module is used to identify whether there are people, obstacles, carriages occupying the lane or tracks in front of the vehicle, and outputs the target category, distance, orientation and corresponding image information.

[0066] 2. Wheel-rail relationship detection module

[0067] The device is used to collect wheel-rail relationship information between the vehicle body and the rails, and preferably includes two or more proximity switches located at the bottom of the vehicle body. The proximity switches can be symmetrically arranged along the left and right sides of the vehicle body and / or spaced apart along the front and rear directions of the vehicle body to improve detection stability and coverage.

[0068] 3. Attitude Detection Module

[0069] The system is used to collect vehicle attitude information, preferably including a tilt sensor. The attitude detection module is used to acquire vehicle roll angle, pitch angle, and related trend information.

[0070] 4. Image verification module

[0071] The system is used to acquire images of the track lines and vehicle body edges and generate image verification information, preferably including a dedicated camera. The image verification module is used to identify the track lines, vehicle body edge lines, and their relative offset relationships to assist in anomaly detection.

[0072] 5. Location and Association Module

[0073] This module is used to acquire vehicle location information, segment information, and historical trajectory information. Preferably, the positioning association module communicates with the mine's existing UWB positioning system to achieve real-time acquisition of vehicle location and segment.

[0074] 6. Data Processing and Control Module

[0075] It is used to perform time synchronization, preprocessing, state quantity extraction, segment adaptive threshold correction, single-channel anomaly detection, multi-source consistency fusion detection, event evidence chain generation, and hierarchical upload control on the data collected by each module.

[0076] The data processing and control module is preferably located in the vehicle control unit or edge processing terminal.

[0077] 7. Power Supply Module

[0078] This power supply provides power to the forward sensing module, wheel-rail relationship detection module, attitude detection module, image verification module, positioning association module, and data processing and control module. An onboard uninterruptible power supply or onboard battery pack is preferred to accommodate situations where the skipcar lacks a fixed power supply.

[0079] 8. Wireless transmission module

[0080] It is used to upload status data, early warning summary data, and event data to the ground platform through underground coal mine safety wireless access points and vehicle-mounted gateways.

[0081] 9. Platform Processing Module

[0082] It is used to receive abnormal status monitoring data and realize functions such as alarm display, historical query, event review, location display, trajectory playback and statistical analysis.

[0083] (iv) Technical solutions for electronic devices and storage media

[0084] The present invention also provides an electronic device, including a processor, a memory, and a program stored in the memory and executable on the processor, wherein the program, when executed by the processor, implements the steps of the above-described method technical solution.

[0085] The present invention also provides a storage medium storing a program, which, when executed by a processor, implements the steps of the above-described method technical solution.

[0086] (v) Beneficial effects

[0087] Compared with the prior art, the present invention has at least the following beneficial effects:

[0088] This invention employs multi-source information, including forward target detection, wheel-rail relationship detection, attitude detection, image verification, and position association, for collaborative judgment. This overcomes the problem of insufficient anti-interference capability of a single sensor in complex downhole environments and improves the reliability of anomaly identification.

[0089] This invention adaptively adjusts the judgment threshold based on the type of the vehicle section, which can take into account the differences in operating characteristics between straight sections and curved sections, and reduce false alarms or missed alarms caused by fixed thresholds.

[0090] This invention distinguishes between instantaneous fluctuations and continuous anomalies by using consistency fusion judgment within a preset time window, which can effectively suppress false alarms caused by vehicle vibration, track unevenness or short-term impact.

[0091] This invention automatically generates an event evidence chain consisting of captured images, key video clips, triggering evidence, location, and trajectory when an anomaly occurs, enhancing the ability to verify and trace abnormal events.

[0092] This invention employs a hierarchical uploading mechanism for status data, early warning summary data, and event data, combined with a cache and retransmission strategy. This allows it to adapt to the actual operating conditions of underground facilities without fixed power supply or network, while also ensuring system real-time performance and long-term operational stability.

[0093] This invention has good engineering application value and promotion value, and is applicable to underground skip car transportation safety monitoring and similar rail transportation scenarios. Attached Figure Description

[0094] Figure 1 This is a block diagram of the overall structure of the underground skip truck abnormal status monitoring system in an embodiment of the present invention.

[0095] Figure 2 This is a schematic diagram of the installation location of the vehicle-mounted sensor in an embodiment of the present invention.

[0096] Figure 3 This is a flowchart of the abnormal state monitoring method in an embodiment of the present invention.

[0097] Figure 4 This is a flowchart illustrating the event evidence chain generation process in an embodiment of the present invention.

[0098] Figure 5 This is a flowchart illustrating the hierarchical upload and cache re-upload process in an embodiment of the present invention.

[0099] in, Figures 1 to 5The same or similar reference numerals in the accompanying drawings indicate the same or similar parts or functional units. Detailed Implementation

[0100] The present invention will be further described below with reference to the accompanying drawings and embodiments. It should be understood that the following embodiments are for illustrative purposes only and are not intended to limit the scope of protection of the present invention.

[0101] Example 1: System Configuration and Implementation

[0102] like Figure 1 and Figure 2 As shown, the underground skip car abnormal state monitoring system in this embodiment includes a forward sensing module, a wheel-rail relationship detection module, an attitude detection module, an image verification module, a positioning association module, a data processing and control module, a power supply module, a wireless transmission module, and a platform processing module.

[0103] The system comprises the following modules: a forward perception module installed at the front of the vehicle to acquire information about targets ahead; a wheel-rail relationship detection module installed at the bottom of the vehicle body to detect the relative position between the vehicle body and the rails; an attitude detection module installed at an appropriate location on the vehicle body to acquire roll and pitch angles; an image verification module installed at a location capable of simultaneously acquiring images of the track line and the vehicle body edges; a positioning association module connected to the mine's existing UWB positioning system; a data processing and control module installed in the onboard control box to synchronize, process, and evaluate multi-source data; a power supply module to power all modules; a wireless transmission module to upload relevant data to the ground platform; and a platform processing module for display and verification.

[0104] Example 2: Implementation Method for Derailment Anomaly Detection

[0105] In this embodiment, the proximity switch continuously detects the relative position between the vehicle body and the rail. When the proximity switch detects abnormal values ​​for multiple consecutive sampling cycles within a certain period, and the tilt sensor detects that the roll angle continuously deviates from the normal range, while the image verification module identifies a significant offset of the vehicle body edge relative to the track line, the system determines within a preset time window that at least two types of abnormal results simultaneously meet the persistence condition, thereby identifying it as a derailment anomaly.

[0106] After a derailment anomaly is determined, the system automatically executes the event evidence chain generation process, which includes capturing key images, saving video clips before and after the anomaly, recording the abnormal results of proximity switches, abnormal attitude results, and abnormal image offset results, and then uploading them to the platform after binding the event location and trajectory information.

[0107] Example 3: Implementation of Rollover Risk Warning

[0108] In this embodiment, if during vehicle operation, the tilt sensor detects a continuous increase in the roll angle, and the image verification results show that the vehicle body gradually deviates from the edge of the track, but the proximity switch results have not yet reached the abnormal event threshold, the system will determine this state as a risk warning rather than directly as an abnormal event.

[0109] In this way, early warnings can be issued before an anomaly develops into an event, prompting relevant personnel to pay attention and investigate in a timely manner.

[0110] Example 4: Implementation of Foreign Object Warning System

[0111] In this embodiment, the forward sensing module continuously monitors the area in front of the vehicle. When the millimeter-wave radar camera detects a person, obstacle, carriage obstructing the aisle, or track obstructing the preset alarm area, the system generates a foreign object alarm event and determines the alarm level based on the target category and distance information.

[0112] After the foreign object alarm is triggered, the system also performs processes such as capturing images, saving videos, binding locations, and extracting trajectories to form a complete chain of event evidence.

[0113] Example 5: Segment Adaptive Threshold Implementation

[0114] In this embodiment, the system determines whether the vehicle is currently in a straight section or a curved section based on the location data provided by the positioning association module. When the vehicle is in a straight section, a first set of thresholds is used; when the vehicle is in a curved section, a second set of thresholds is used. In the second set of thresholds, a larger normal fluctuation range is allowed for roll angle, pitch angle, and image offset.

[0115] By adjusting the threshold according to the section, false alarms during curve operation can be effectively reduced and the overall judgment accuracy can be improved.

[0116] Example 6: Implementation of Tiered Upload and Cache Re-upload

[0117] In this embodiment, when the system is running normally, it only uploads status data periodically; when a risk warning is detected, the warning level and trend summary information are uploaded; when an abnormal event or foreign object alarm event is detected, key images, triggering evidence and location information are uploaded first, and key video clips are uploaded when bandwidth allows.

[0118] When the underground wireless network is interrupted, the system caches event data to local storage; once the network is restored, it prioritizes retransmitting data related to abnormal events to ensure that critical events are not lost.

[0119] Example 7: Alternative Implementation

[0120] In other embodiments, the wheel-rail relationship detection module may use other distance detection or proximity detection elements suitable for the underground environment, in addition to proximity switches; the attitude detection module may use an inertial measurement unit in addition to tilt sensors to obtain attitude information; the forward sensing module may use an integrated or separate structure of radar and camera; and the positioning association module may be connected to other mine vehicle positioning systems in addition to UWB positioning systems.

[0121] It should be understood that any equivalent substitutions or modifications made to the form of each module, installation method, parameter settings, judgment strategy, and upload strategy without departing from the concept of this invention shall fall within the protection scope of this invention.

Claims

1. A method for monitoring abnormal states of underground skip cars based on multi-source sensing fusion, characterized in that, Includes the following steps: S1 collects information on targets in front of the vehicle, wheel-rail relationship information, vehicle attitude information, image verification information, and vehicle position information; S2 performs time synchronization, preprocessing, and state quantity extraction on the collected information to obtain multi-source state data under a unified time reference. S3, based on the vehicle location information, identify the current segment type of the vehicle, and adaptively correct the wheel-rail distance threshold, attitude threshold, and image offset threshold based on the segment type; S4, obtain wheel-rail relationship anomaly results based on the wheel-rail relationship information, obtain attitude anomaly results based on the vehicle body attitude information, and obtain image offset anomaly results based on the image verification information; S5, within a preset time window, perform consistency fusion judgment on the abnormal results of wheel-rail relationship, abnormal results of attitude and abnormal results of image offset. When at least two types of abnormal results meet the duration condition or the continuous frame condition, they are judged as abnormal events. When the criteria for determining an abnormal event are not met, but at least one type of abnormal result meets the condition of a continuous abnormal change trend, it is judged as a risk warning; S6, identify the target information ahead, and generate a foreign object alarm event when at least one of the following targets, such as personnel, obstacles, carriages occupying the road or tracks, enters the preset alarm area; S7. When the abnormal event or the foreign object alarm event is determined, the following actions are triggered: snapshot capture, key video clip saving, sensor triggering data recording, location information binding, and historical trajectory extraction to form an event evidence chain. S8 uploads the corresponding data according to the hierarchical upload strategy. Under normal conditions, it uploads status data; under risk warning conditions, it uploads warning summary data; and when an abnormal event or foreign object alarm event is triggered, it uploads captured images, key video clips, triggering basis, event location and trajectory information.

2. The method for monitoring abnormal conditions of underground skip cars according to claim 1, characterized in that, The wheel-rail relationship information is obtained by two or more proximity switches located at the bottom of the vehicle body. The proximity switches are symmetrically arranged along the left and right sides of the vehicle body and / or spaced apart along the front and rear directions of the vehicle body.

3. The method for monitoring abnormal conditions of an underground skip car according to claim 1, characterized in that, The vehicle attitude information includes roll angle information and pitch angle information, and the attitude anomaly results are obtained through at least one of attitude angle threshold judgment and attitude change rate judgment.

4. The method for monitoring abnormal conditions of an underground skip car according to claim 1, characterized in that, The image verification information is obtained by identifying the track boundary line and the vehicle body edge line, and the image offset anomaly result is obtained by calculating the offset of the vehicle body edge relative to the track boundary line.

5. The method for monitoring abnormal conditions of an underground skip as described in claim 1, characterized in that, The segment types include at least straight segments and curved segments, and different segment types correspond to different wheel-rail distance thresholds, attitude thresholds, and image offset thresholds.

6. The method for monitoring abnormal conditions of an underground skip car according to claim 1, characterized in that, The consistency fusion judgment includes time consistency judgment and persistence consistency judgment; the time consistency judgment is used to determine whether at least two types of abnormal results meet the overlap condition within the same preset time window, and the persistence consistency judgment is used to determine whether the abnormal results continue to reach a set duration or number of consecutive frames.

7. The method for monitoring abnormal conditions of an underground skip car according to claim 1, characterized in that, The conditions for the continuous abnormal change trend include at least one of the following: The wheel-rail relationship test results continuously deviated from the normal range and showed an increasing trend; The roll or pitch angle continuously deviates from the normal range and shows an increasing trend; The image offset continued to increase and did not return to the normal range.

8. The method for monitoring abnormal conditions of an underground skip car according to claim 1, characterized in that, The event evidence chain shall include at least one of the following: event number, event time, event type, triggering basis, captured image, key video clip, event location, segment to which it belongs, and historical trajectory.

9. The method for monitoring abnormal conditions of an underground skip car according to claim 1, characterized in that, The tiered upload strategy includes a local caching and retransmission mechanism, which caches abnormal event data locally when the wireless network is interrupted and retransmits it preferentially after the network is restored.

10. A method for monitoring abnormal states of underground skip cars based on multi-source sensing fusion, characterized in that, include: The forward perception module is used to collect information about targets in front of the vehicle; The wheel-rail relationship detection module is used to collect wheel-rail relationship information between the car body and the rail; The attitude detection module is used to collect vehicle attitude information; The image verification module is used to acquire images of the track line and vehicle body edges and obtain image verification information; The location association module is used to obtain vehicle location information, segment information, and historical trajectory information; The data processing and control module is used to perform time synchronization, preprocessing, state quantity extraction, segment adaptive threshold correction, consistency fusion judgment, event evidence chain generation, and hierarchical upload control on the data collected by each module. The power supply module is used to supply power to the forward sensing module, wheel-rail relationship detection module, attitude detection module, image verification module, positioning association module and data processing and control module; The wireless transmission module is used to upload status data, early warning summary data, and event data to the ground platform; The platform processing module is used for alarm display, historical query, event review, and trajectory playback; The data processing and control module is configured to perform the method described in any one of claims 1 to 9.

Images