An underground space safety situation edge data visualization method and system

Abstract

The application relates to the technical field of image processing, and discloses an underground space safety situation edge data visualization method and system. The method comprises the following steps: performing dynamic target detection on standardized data of an underground space to obtain initial target information of the underground space; based on multi-view geometric constraints of the underground space, the motion consistency of the initial target information under different fields of view is evaluated to establish preliminary cross-view trajectory correlation of the underground space; according to the preliminary cross-view trajectory correlation, the initial target information is fused into continuous trajectories under a unified coordinate system to generate fusion trajectory data of the underground space; the fusion trajectory data is projected to a three-dimensional space, and natural object falling traceability analysis is performed on the fusion trajectory data to obtain traceability position information of the underground space; the traceability position information and a preset safety knowledge graph are integrated to generate safety situation visualization data of the underground space; and the application can improve the visualization efficiency of underground space safety situation edge data.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and in particular to a method and system for visualizing edge data of underground space security situation. Background Technology

[0002] Underground spaces are enclosed environments with complex structures, and their safety monitoring relies on multi-source, heterogeneous edge data. Current technologies lack efficient standardized processing mechanisms in the data preprocessing stage, making it difficult to accurately unify the format of multi-source monitoring data, remove redundancy, and complete key fields. This results in low efficiency in extracting effective data, creating potential risks for subsequent safety situation analysis. Furthermore, in the dynamic target detection and cross-view trajectory association stages, existing technologies do not fully integrate the actual structural layout of underground spaces to construct multi-view geometric constraints. The assessment of the consistency of target movement under different views lacks scientific quantitative basis, leading to broken or erroneous cross-view trajectory associations and preventing the formation of complete target movement trajectories.

[0003] Existing technologies have significant shortcomings in trajectory fusion and visualization. They fail to establish a unified spatial coordinate system for accurate calibration and connection of scattered trajectory segments, resulting in poor continuity and significant spatial positioning deviations in the fused trajectory data. Furthermore, they lack systematic methods for tracing the source of safety risks such as falling objects, making it difficult to accurately pinpoint the location of the risk source. They also fail to achieve deep integration of source information with safety knowledge graphs, lacking sufficient multimodal conversion capabilities for core safety information. The intuitiveness and practicality of the visualized data are also poor, failing to provide efficient support for real-time monitoring and risk warning of underground space safety situations, leading to delayed identification of safety hazards and untimely response. Summary of the Invention

[0004] This invention provides a method and system for visualizing edge data on the safety situation of underground spaces, in order to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides a method for visualizing edge data of underground space safety situation, comprising: S1. Obtain standardized data on underground space; S2. Perform dynamic target detection on the standardized data to obtain the initial target information of the underground space; S3. Based on the multi-view geometric constraints of the underground space, evaluate the motion consistency of the initial target information under different viewpoints to establish a preliminary cross-view trajectory association of the underground space; S4. Based on the preliminary cross-view trajectory association, the initial target information is fused into a continuous trajectory under a unified coordinate system to generate the fused trajectory data of the underground space; S5. Project the fused trajectory data into a three-dimensional space, and perform a natural object fall source tracing analysis on the fused trajectory data to obtain the source location information of the underground space; S6. Integrate the source location information with the preset safety knowledge graph to generate visual data on the safety situation of the underground space.

[0006] In a preferred embodiment, acquiring standardized data of underground space includes: Acquire multi-source monitoring data of underground space; The multi-source monitoring data is processed to unify the format, resulting in unified format data of the underground space. Redundant and invalid data are removed from the standardized data to obtain valid data for the underground space. By completing the key data fields of the valid data, the standardized data of the underground space is obtained.

[0007] In a preferred embodiment, the step of performing dynamic target detection on the standardized data to obtain initial target information of the underground space includes: Based on the safety monitoring requirements of the underground space, the dynamic target detection range of the underground space is defined; Target recognition is performed on the image data in the standardized data to capture target objects with motion characteristics in the underground space; The target object's morphological characteristics, time of appearance, and initial location in the monitoring field of view are integrated into basic attribute information; The basic attribute information is classified and organized to obtain the initial target information of the underground space.

[0008] In a preferred embodiment, the step of evaluating the motion consistency of the initial target information under different viewpoints based on the multi-view geometric constraints of the underground space to establish a preliminary cross-view trajectory association of the underground space includes: Based on the actual structural layout of the underground space, the relative positional relationships of the multi-view monitoring devices in the underground space are determined in order to establish the multi-view geometric constraints of the underground space. Identify the movement direction and rhythm trend of the same target object in the initial target information under different monitoring views to obtain the movement characteristics of the initial target information; Based on the multi-view geometric constraints, the degree of fit of the motion features under different monitoring views is evaluated to obtain the motion consistency of the initial target information; The target motion trajectory segments in the initial target information that meet the preset conditions for motion consistency are associated and matched to establish a preliminary cross-view trajectory association for the underground space.

[0009] In a preferred embodiment, the formula for calculating the degree of fit is as follows: ; In the formula, Indicates the degree of fit. This indicates the number of samples within a preset time window. The viewpoint in the multi-view geometric constraint condition represents the perspective. To view The fundamental matrix, Indicates time From the perspective of time The target location coordinates in the middle, Indicates time From the perspective of time The target location coordinates in the middle, This represents the preset scaling parameter used to normalize the polar error. An exponential function representing the natural constant.

[0010] In a preferred embodiment, the step of fusing the initial target information into a continuous trajectory in a unified coordinate system based on the preliminary cross-view trajectory association to generate the fused trajectory data of the underground space includes: A unified spatial coordinate system for the underground space is established by using a pre-set fixed reference marker within the underground space as a reference. The trajectory segments in the preliminary cross-view trajectory association are mapped to the unified coordinate system to calibrate the spatial position of the trajectory segments, thereby obtaining the calibrated trajectory segments of the underground space; Determine the rationality of the connection between the trajectory segments in order to retain the trajectory connection relationships with high consistency among the trajectory segments; Based on the trajectory connection relationship, reasonable transition information at the connection point of the calibration trajectory segment is supplemented to obtain the fused trajectory data of the underground space.

[0011] In a preferred embodiment, the step of projecting the fused trajectory data into three-dimensional space and performing natural object fall source tracing analysis on the fused trajectory data to obtain the source location information of the underground space includes: Based on the actual three-dimensional structural data of the underground space, determine the correspondence rules between the fused trajectory data and the physical location in the three-dimensional space; Based on the correspondence, the fused trajectory data is projected onto the three-dimensional space to obtain the three-dimensional trajectory data of the underground space; By filtering the three-dimensional trajectory data to obtain the trajectory fragments of the natural objects, the trajectory data corresponding to the natural objects are obtained. Based on the trajectory supplementary information recorded by the multi-view monitoring equipment in the underground space, the trajectory deviation of the natural object movement trajectory segment is corrected; By tracing the starting point of the corrected trajectory of the natural object and combining it with the position markers in the three-dimensional space, the starting spatial coordinates of the natural object's fall are determined. Based on the specific regional division of the underground space, the spatial positioning verification of the starting spatial coordinates is performed to obtain the source location information of the underground space.

[0012] In a preferred embodiment, the step of integrating the source location information with a preset safety knowledge graph to generate visualized safety situation data for the underground space includes: Based on the spatial identifier in the source location information, match the corresponding spatial region and associated security rules in the preset security knowledge graph; Based on the safety rules, the types of natural objects corresponding to the source location information are associated with the types of hazards in the safety knowledge graph to determine the potential safety risk attributes of the underground space; Verify the logical consistency between the source location information and the potential security risk attributes to correct the source location information; Multimodal evolution is performed on the core security information in the corrected source location information to obtain the security situation visualization data of the underground space.

[0013] In a preferred embodiment, the step of performing multimodal evolution on the core security information in the corrected source location information to obtain the security situation visualization data of the underground space includes: According to the preset multimodal presentation specifications, the text-based association information of the core security elements in the corrected source location information is converted into standardized explanatory text to obtain the standardized text data of the corrected source location information. Using preset graphical identification rules, the spatial association data in the core safety elements is graphically transformed to obtain the spatial graphical data of the underground space; Add time-series association markers to the standardized text data and the spatial graphical data to bind the state change trend of core security information with the time dimension, thereby obtaining the multimodal integrated data of the underground space; Based on the requirements of the underground space security situation display scenario, duplicate information elements in the multimodal integrated data are removed, and the information complementarity and logical coherence between different modal data are strengthened to obtain the underground space security situation visualization data.

[0014] To address the aforementioned problems, the present invention also provides an edge data visualization system for underground space safety situation, the system comprising: The data acquisition module is used to acquire standardized data of underground space; The target detection module is used to perform dynamic target detection on the standardized data to obtain the initial target information of the underground space; The trajectory association module is used to evaluate the motion consistency of the initial target information under different viewpoints based on the multi-view geometric constraints of the underground space, so as to establish a preliminary cross-view trajectory association of the underground space. The trajectory fusion module is used to fuse the initial target information into a continuous trajectory in a unified coordinate system based on the preliminary cross-view trajectory association, so as to generate the fused trajectory data of the underground space; The source tracing analysis module is used to project the fused trajectory data into a three-dimensional space and perform a natural object fall source tracing analysis on the fused trajectory data to obtain the source location information of the underground space; The visualization generation module is used to integrate the source location information with the preset safety knowledge graph to generate visualization data of the safety situation of the underground space.

[0015] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention transforms multi-source monitoring data into standardized data through format unification, redundancy removal, and key field completion, solving the problems of data disorder and difficulty in extracting effective information in existing technologies. Furthermore, it assesses the consistency of target motion based on multi-view geometric constraints in underground space, and combines unified coordinate system calibration and connection of trajectory segments to avoid cross-view trajectory breaks or miscorrelation, generating continuous fused trajectory data, providing a high-quality data foundation for subsequent safety analysis.

[0016] 2. This invention accurately traces the starting position of falling natural objects by projecting the fused trajectory into three-dimensional space and correcting trajectory deviations by combining supplementary information from multiple perspectives. It also identifies potential risk attributes by matching a safety knowledge graph and transforms core safety information into text and graphic data with temporal correlations through multimodal evolution. This solves the problems of inaccurate tracing and poor visualization in existing technologies, and can provide clear support for real-time monitoring and risk warning of underground space safety, thereby improving safety management efficiency. Attached Figure Description

[0017] Figure 1 A flowchart illustrating a method for visualizing edge data on the safety situation of underground space, provided in an embodiment of the present invention; Figure 2 A functional module diagram of an underground space safety situation edge data visualization system provided in an embodiment of the present invention; The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0018] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0019] This application provides a method for visualizing edge data of underground space security situation. The executing entity of this method includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this application: a server, a terminal, etc. In other words, the method for visualizing edge data of underground space security situation can be executed by software or hardware installed on a terminal device or a server device. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster. The server can be an independent server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0020] Reference Figure 1 The diagram shown is a flowchart illustrating a method for visualizing edge data on the safety situation of underground space according to an embodiment of the present invention. In this embodiment, the method for visualizing edge data on the safety situation of underground space includes: S1. Obtain standardized data on underground space; In this embodiment of the invention, acquiring standardized data of underground space includes: Acquire multi-source monitoring data of underground space; The multi-source monitoring data is processed to unify the format, resulting in unified format data of the underground space. Redundant and invalid data are removed from the standardized data to obtain valid data for the underground space. By completing the key data fields of the valid data, the standardized data of the underground space is obtained.

[0021] The process of acquiring multi-source monitoring data for underground spaces revolves around the key monitoring needs of these spaces. This involves collecting relevant data through various monitoring devices deployed in different areas of the underground space. These devices include geological structure monitoring equipment, environmental sensors, and spatial imaging equipment. Geological structure monitoring equipment collects data such as rock strata displacement and geological stress. Environmental sensors capture environmental parameters such as temperature, humidity, and gas concentration within the underground space. Spatial imaging equipment acquires three-dimensional morphological images and detailed images of the internal structure of the underground space. During data acquisition, a stable connection is established between various monitoring devices and the data acquisition terminal via wired transmission lines. The monitoring devices continuously collect raw data at a preset acquisition frequency. The collected raw data undergoes preliminary signal amplification and noise reduction processing through the device's built-in signal processing module to avoid signal distortion during transmission. The processed raw data is then transmitted to the data acquisition terminal in real time via the transmission lines. The data acquisition terminal records each received raw data point, including the specific location of the monitoring area, the acquisition time, and the original monitoring information. For numerical data, the original monitoring information directly records the measured physical quantity values; for image data, it records complete pixel array information and imaging parameters. During the recording process, the data acquisition terminal performs real-time verification to ensure that the received data is neither lost nor duplicated, ultimately forming a multi-source monitoring data set covering all dimensions of underground space monitoring, completely preserving various original monitoring information, and providing a foundation for subsequent processing.

[0022] When standardizing the format of multi-source monitoring data, the first step is to clarify the differences in the original formats of the data. Different types of monitoring equipment output data in different formats. Geological structure monitoring equipment may output numerical data in text format, environmental sensors may output parameter data in binary format, and space imaging equipment may output image data with different resolutions and encoding formats. The format standardization process first performs targeted parsing for each original data format. For text-format numerical data, the monitoring values ​​and related information are extracted by reading line by line, and irrelevant annotations are removed. For binary-format parameter data, the binary code is converted into readable decimal values ​​and related descriptive information according to the preset decoding rules of the corresponding monitoring equipment. For image data, the core data such as pixel information, imaging size, and color channels are extracted by parsing the output protocol of the imaging equipment, and redundant equipment identification information is removed. Subsequently, a unified data format standard was established, clarifying the arrangement order of data fields, including a fixed sorting of fields such as monitoring location, acquisition time, monitoring type, monitoring values, and core parameters. Numerical data was uniformly represented in decimal pure numeric form, time information was uniformly represented in continuous character combinations, and image data was uniformly set to a fixed resolution and color encoding format. Fields were distinguished by uniform character separators. The parsed data was reorganized according to the established unified format standard, and the field information of each data point was checked to ensure that the content of each field corresponded accurately. Numerical data was adjusted to a unified representation format, time information was converted to a standard format, and image data was adjusted and re-encoded according to the preset resolution and encoding format. After reorganization, the format of each data point was verified to confirm that there were no field misalignments or format inconsistencies, ultimately resulting in unified format data for underground space.

[0023] When removing redundant and invalid data from the standardized data, the criteria for determining redundant and invalid data must first be clearly defined. Redundant information refers to identical data repeatedly recorded in the standardized data and additional information unrelated to the underground space monitoring target. Invalid data includes distorted data generated during the transmission of abnormal numerical data beyond the reasonable monitoring range and incomplete data lacking core information. During processing, the standardized data is first traversed line by line. For duplicate data, the monitoring location, acquisition time, core monitoring values, or key image features of each data point are compared. If all core information is completely identical, it is determined to be duplicate data, and the first data point to appear is retained, while subsequent identical data are removed. For additional information unrelated to the monitoring target, based on the core requirements of underground space monitoring, key information related to the geological structure, environmental state, and spatial morphology is selected, while additional content such as identifier characters unrelated to equipment operation logs is removed. For invalid data, based on the reasonable data range for different monitoring types, such as the normal range of geological stress, the reasonable fluctuation range of temperature and humidity, and the effective pixel ratio standard of image data, values ​​exceeding the range or image data that do not meet the standard are judged as abnormal data and removed. By checking the integrity of the data, if key fields such as monitoring location, acquisition time, and core monitoring values ​​are missing, or if the image data has a large area of ​​pixel loss and the core information cannot be identified, it is judged as incomplete data and removed. By comparing continuous monitoring data at the same monitoring location, if the value of a certain data point has a significant abrupt change with the data in adjacent time periods without a reasonable cause, or if the characteristics of the image data are obviously contradictory to the characteristics of normal images in the same area, it is judged as distorted data and removed. During the removal process, each piece of data that is removed is recorded separately, clearly defining the removal type and the judgment basis to ensure that the removal operation is traceable, and finally obtains the valid data of underground space.

[0024] When supplementing key data fields in valid data, first clarify the specific content of the key data fields, including a detailed description of the underground space monitoring location, complete information on the collection time, the unique identifier of the monitoring equipment, and the specific values ​​of core monitoring indicators or core image parameters. These fields are the foundation for ensuring data usability and the effectiveness of subsequent analysis. During processing, each valid data point is checked individually to verify the completeness of each key data field, identifying any missing fields that need to be supplemented. For data lacking detailed descriptions of monitoring locations, the specific installation location information of the equipment in the underground space, including horizontal area division and vertical depth information, is obtained by querying the equipment deployment files recorded by the data acquisition terminal and based on the monitoring equipment number corresponding to the data. This information is then added to the monitoring location field to ensure accurate and detailed descriptions. For data lacking acquisition time, the accurate acquisition time is calculated based on the time record of the data received by the data acquisition terminal and the average signal transmission delay time between the monitoring equipment and the acquisition terminal. This time information is then added to the acquisition time field according to a unified time format. For data lacking the unique identifier of the monitoring equipment, the corresponding unique identifier information is obtained and completed by querying the equipment management file using the equipment communication address recorded during data transmission. For data lacking core monitoring indicator values ​​or core image parameters, if other valid data from the same acquisition period exists for the corresponding monitoring location, the missing core information is supplemented by calculating the average value of adjacent valid data or extracting core parameter features from similar images. If no data from the same period exists, historical monitoring data for the monitoring location is queried, data change patterns are analyzed, and the missing core field content is reasonably deduced by combining this with the overall monitoring trend of the area within the current monitoring cycle. After the data is completed, the key fields of each data entry are re-verified to ensure that all key data fields are complete and without missing information, and that the completed information is consistent with the overall characteristics of the data and has no logical contradictions. This results in the final standardized data for underground space.

[0025] For example, in a subway tunnel safety monitoring project in a certain city, the system collects raw multi-source data through various monitoring devices deployed in the underground space: geological structure monitoring devices record rock strata displacement in text format, with a sampling frequency of 1 time / second, and data fields include "monitoring point number, timestamp, lateral displacement, and longitudinal displacement"; environmental sensors output in binary format, collecting data once every 5 seconds, with content including "temperature, humidity, and methane concentration"; and network high-definition cameras transmit 1920×1080 resolution video streams in real time with H.264 encoding at a frame rate of 25fps.

[0026] First, the format was standardized: a Python script was used to parse the text data, extract key fields, and convert it into JSON format; the binary data was converted into a CSV table with timestamps using the decoding library provided by the device manufacturer; and the video stream was uniformly transcoded into a standard format with a resolution of 1920×1080, a frame rate of 30fps, and H.265 encoding using FFmpeg.

[0027] Then, redundant and invalid data were removed: the Pandas library was used to deduplicate temperature data within 10 consecutive minutes, and duplicate records with identical values ​​were deleted; a reasonable threshold for rock stratum displacement (lateral displacement ≤ 50 mm) was set based on engineering experience, and abnormal data exceeding this threshold were removed; empty frames without moving targets in the video frame were identified and removed using the background subtraction method of OpenCV.

[0028] Finally, key data fields were completed: By querying the device management database, the installation location description of the corresponding camera was completed for video frames missing the "monitoring location" field, such as "K12+350 south wall-mounted point on the left tunnel line"; based on the NTP synchronization time server and transmission delay model, the accurate acquisition timestamps missing from some sensors were calculated and completed; for individual records missing "methane concentration" values, linear interpolation of data from the same sensor within 5 minutes before and after was used to complete the data. The final output is a standardized dataset with a well-structured, complete, and time-synchronized dataset, stored in the time-series database InfluxDB, providing high-quality input for subsequent analysis.

[0029] The beneficial effects are that the multi-source monitoring data processing process implemented in stages achieves standardized processing of underground space monitoring data from acquisition to standardization. The format unification process eliminates the format differences of multi-source data, ensuring data consistency and compatibility. The removal of redundant information and invalid data improves data quality. The completion of key data fields ensures data integrity and usability. The resulting standardized data provides high-quality data support for subsequent analysis and evaluation of underground space, improving the utilization efficiency and analysis accuracy of underground space monitoring data.

[0030] S2. Perform dynamic target detection on the standardized data to obtain the initial target information of the underground space; In this embodiment of the invention, the step of performing dynamic target detection on the standardized data to obtain initial target information of the underground space includes: Based on the safety monitoring requirements of the underground space, the dynamic target detection range of the underground space is defined; Target recognition is performed on the image data in the standardized data to capture target objects with motion characteristics in the underground space; The target object's morphological characteristics, time of appearance, and initial location in the monitoring field of view are integrated into basic attribute information; The basic attribute information is classified and organized to obtain the initial target information of the underground space.

[0031] Based on the safety monitoring needs of underground spaces, the scope of dynamic target monitoring is defined, and the core requirements for underground space safety monitoring are comprehensively reviewed, including specific aspects such as preventing personnel from accidentally entering unauthorized dangerous areas, monitoring abnormal movement of key equipment, identifying foreign objects intruding into core passages, and monitoring suspicious activities around pipelines. Combining the physical structure and layout of the underground space, key safety protection areas are identified through on-site surveys, covering main passage entrances and exits, equipment rooms and surrounding areas, densely packed pipeline sections, geologically unstable areas, and emergency evacuation routes. These areas are the core areas that dynamic target detection must prioritize. The physical boundaries of each key protection area are determined. Horizontally, fixed structures such as walls, columns, and railings within the area are used as the basis for defining the coverage boundaries in the left-right and front-back directions. Vertically, based on the actual height of the underground space, a complete vertical detection range is delineated from the ground upwards to below the top structure to ensure no blind spots. Simultaneously, based on security monitoring requirements, the types of dynamic targets that need to be detected are clearly defined, including personnel, various mobile devices, and unidentified foreign objects. The physical scope of key protection areas is then mapped to the target types. For example, the personnel detection scope covers all passageways and unauthorized entry areas, the equipment detection scope focuses on the equipment room and surrounding areas, and the foreign object detection scope covers critical passageways and the area around pipelines. This ultimately forms a clear dynamic target detection scope, which clarifies the specific spatial area requiring dynamic target monitoring and the corresponding target types to be detected, providing clear scope guidance for subsequent target identification work.

[0032] Target recognition is performed on image data from standardized data sets to capture moving objects in underground spaces. All image data, with uniform resolution and color encoding formats, is extracted from the standardized data set to ensure consistency in the recognition process. The extracted image data is arranged sequentially according to the acquisition time to form a continuous image sequence, ensuring temporal continuity between image frames. Pixel distribution features are extracted frame by frame for each image. By scanning each pixel, the color value of each pixel and the positional relationships between adjacent pixels are recorded, constructing a complete pixel feature set for each frame. The pixel feature sets of two consecutive frames are compared, and pixels at corresponding positions are examined one by one to identify pixel regions where color values ​​or positional relationships have changed. These changed regions are temporarily marked. The changes of the temporarily marked regions in subsequent consecutive frames are tracked to observe whether they maintain a complete outline shape and exhibit a continuous movement trajectory, thus eliminating local pixel changes caused by non-moving target factors such as changes in light intensity or floating dust particles. For marked regions that meet the characteristics of continuous movement and have complete outlines, their specific range in the image is determined by delineating the edge pixels of the region. This confirms that the entity corresponding to the region is a target object with motion characteristics. All frames in the image sequence are processed one by one according to the above process to completely capture all target objects with motion characteristics appearing in the underground space.

[0033] The initial monitoring field of view at the moment the target object's morphological features appeared is integrated into the basic attribute information. For each identified target object, its morphological features are extracted from the first clearly identified image frame. By scanning the image region corresponding to the target object point by point, the outer contour shape of the target object is delineated by connecting the edge pixels. The total number of image pixels occupied by this contour is counted to determine the size of the target object. At the same time, the color values ​​of all pixels in the target object region are extracted, and the main color distribution characteristics of the target object are determined after summarizing and analyzing. The contour shape, size, and color distribution are integrated into a complete target morphological feature. The acquisition time corresponding to the first image frame of the target object is retrieved from the standardized data. This time is the moment the target object appeared, ensuring that the time information is completely consistent with the image data acquisition record without deviation. By querying the unique identifier of the monitoring equipment that acquired the image data, combined with the physical coverage area information corresponding to the identifier recorded in the equipment deployment file, the monitoring field of view at the time the target object was first identified is determined. This field of view information includes the area name and specific physical location range of the monitoring equipment. Each target object is uniquely associated with three pieces of information: its morphological characteristics, the time of its appearance, and its initial location within the monitoring field of view. This ensures that each piece of information accurately corresponds to a specific target object, without any confusion or mismatch, ultimately forming the unique basic attribute information for each target object.

[0034] The initial target information for underground space is obtained by classifying and organizing the basic attribute information. Based on the safety monitoring needs of underground space and the characteristics of common dynamic targets, clear classification criteria are formulated. Target objects are divided into three categories: personnel, mobile devices, and foreign objects. The classification criteria for personnel targets are: an upright outline with distinct head, torso, and limb outlines; a size corresponding to the image pixel range of human height; and a diverse and irregular color distribution. The classification criteria for mobile devices targets are: a regular geometric shape; a size larger than the image pixel range corresponding to small objects; and a relatively uniform color distribution, mostly metallic or a fixed color common to industrial equipment. The classification criteria for foreign objects targets are: an irregular shape; a size exceeding the common image pixel range of personnel and mobile devices; or an outline feature with no similarity to personnel or mobile devices. The target morphological features in each basic attribute information are extracted and compared item by item with the formulated classification criteria. The specific category of the target object is determined based on the comparison results. Basic attribute information of target objects of the same category is grouped together, with each group clearly labeled with its corresponding category name. During the grouping process, each basic attribute information is double-checked to ensure accurate category identification and prevent cross-category misplacement. The basic attribute information within each group is then arranged in chronological order of appearance, forming a structured set of category information. By summarizing all the structured information sets of all categories, the initial target information for underground space is obtained. This information comprehensively includes the basic attributes and classification results of various dynamic targets, providing clear data support for subsequent underground space safety status analysis.

[0035] Based on standardized video data, the system defines the dynamic target detection range according to the "Regulations for Safety Monitoring of Metro Tunnels": horizontally, it covers the main passage area of ​​the tunnel (3 meters to the left and right of the track centerline), the equipment installation area (2 meters around the power distribution box and fan), and emergency escape routes; vertically, it extends from the track surface to the tunnel arch (height range 0-6.5 meters). Target types are defined into three categories: personnel (wearing reflective clothing or regular clothing), engineering vehicles (tunnel inspection vehicles, material transport vehicles), and potential falling objects (concrete blocks, bolts, etc. that have fallen from the tunnel wall or ceiling).

[0036] Real-time target recognition was performed using the YOLOv5s model, which was jointly trained on the COCO dataset and a self-built tunnel scene dataset. The input image size was 640×640, and the detection threshold was set to 0.5. In processing a 30-second continuous video stream, the system identified a target at frame 125 (14:30:22), with bounding box coordinates of [320, 150, 420, 300] and a class confidence score of 0.87, which was determined to be an "engineering vehicle".

[0037] The target's morphological features were extracted: Edge detection and contour analysis yielded a bounding rectangle with dimensions of 6.2 meters and 2.4 meters (calculated based on pixel size and camera calibration parameters). The color histogram showed the dominant color as yellow (H channel peak value in HSV space between 50-60). Its basic attributes were recorded: Occurrence time was 2023-10-10 14:30:22.500; initial monitoring field of view number was Cam_03 (corresponding to location K2+100 in the southern section of the tunnel).

[0038] Finally, based on the classification rules: personnel targets have human-like outlines and a height between 1.5 and 2.0 meters; vehicle targets have regular geometric shapes and a length greater than 3 meters; falling objects usually have irregular shapes and their trajectories exhibit free-fall characteristics, the system classifies the target as "mobile device" and integrates all its attributes into a structured record, storing it in the target information table.

[0039] The beneficial effects include: accurately defining the detection range of dynamic targets based on the needs of underground space safety monitoring; clarifying key protection areas and corresponding monitoring target types; avoiding redundancy or omissions in the detection range; providing clear guidance for subsequent target identification; and ensuring that monitoring work focuses on core safety requirements. The target identification process for standardized image data accurately captures moving targets by comparing pixel features and tracking motion trajectories across consecutive image frames, effectively eliminating interference from non-motion factors and ensuring the accuracy and completeness of target identification. The integration of target morphological characteristics, time of appearance, and initial monitoring field of view into basic attribute information comprehensively records key initial information of the target object, ensuring that each piece of information is uniquely bound to the target object, providing complete data support for subsequent classification and organization. Classifying and organizing the basic attribute information according to clear classification standards makes the initial target information structured and clearly categorized, facilitating rapid differentiation of different types of dynamic targets. This significantly improves the readability and usability of dynamic target information in underground spaces, providing a precise and organized data foundation for subsequent work such as underground space safety status analysis and risk warning, and strengthening the targeting and effectiveness of underground space safety monitoring.

[0040] S3. Based on the multi-view geometric constraints of the underground space, evaluate the motion consistency of the initial target information under different viewpoints to establish a preliminary cross-view trajectory association of the underground space; In this embodiment of the invention, the step of evaluating the motion consistency of the initial target information under different viewpoints based on the multi-view geometric constraints of the underground space, in order to establish a preliminary cross-view trajectory association of the underground space, includes: Based on the actual structural layout of the underground space, the relative positional relationships of the multi-view monitoring devices in the underground space are determined in order to establish the multi-view geometric constraints of the underground space. Identify the movement direction and rhythm trend of the same target object in the initial target information under different monitoring views to obtain the movement characteristics of the initial target information; Based on the multi-view geometric constraints, the degree of fit of the motion features under different monitoring views is evaluated to obtain the motion consistency of the initial target information; The target motion trajectory segments in the initial target information that meet the preset conditions for motion consistency are associated and matched to establish a preliminary cross-view trajectory association for the underground space.

[0041] The formula for calculating the degree of fit is as follows: ; In the formula, Indicates the degree of fit. This indicates the number of samples within a preset time window. The viewpoint in the multi-view geometric constraint condition represents the perspective. To view The fundamental matrix, Indicates time From the perspective of time The target location coordinates in the middle, Indicates time From the perspective of time The target location coordinates in the middle, This represents the preset scaling parameter used to normalize the polar error. An exponential function representing the natural constant.

[0042] The relative positions of multi-view monitoring devices were determined by combining the actual structural layout of the underground space to establish multi-view geometric constraints. Complete physical structural information of the underground space was obtained through on-site surveys, including key structural details such as passageway orientation, wall distribution, column locations, floor height, and area boundaries. The actual installation positions of all deployed monitoring devices were recorded simultaneously, clarifying the installation carrier, height, and orientation angle of each device. Using the fixed structure at the entrance of the underground space as a reference point, a spatial coordinate system was established along both horizontal and vertical directions. Measurement tools were used to determine the specific coordinates of each monitoring device within this reference system, and the straight-line distance between any two monitoring devices was measured, recording their relative orientation. The lens parameters and installation angle of each monitoring device were analyzed to determine its monitoring field of view coverage, shooting angle, and effective monitoring distance. Overlapping and connecting areas between the fields of view of different monitoring devices were identified, and adjacent device combinations with non-overlapping but spatially continuous fields of view were marked. Based on the relative position coordinates, distances, orientational relationships, and field-of-view coverage characteristics of the devices obtained above, multi-view geometric constraints are established. These constraints clearly define the spatial position association rules between different monitoring devices, including the reasonable path range for a target object to move from one device's field of view to another, the scaling relationship of target imaging in different fields of view, and the corresponding position mapping rules of targets in overlapping areas. This ensures that the constraints can accurately reflect the matching relationship between the underground space physical structure and the layout of monitoring devices, providing a spatial basis for subsequent motion characteristic assessment.

[0043] The movement characteristics of the initial target information are obtained by identifying the movement direction and rhythm trend of the same target object under different monitoring views. All basic attribute information containing the same target object is filtered out from the initial target information. Based on the morphological characteristics and chronological order of the target object's appearance, information entries of the target object recorded under different monitoring views are associated. For each target object under each monitoring view, its position information in consecutive image frames is extracted. By comparing the center coordinates of the pixel regions corresponding to the target object in adjacent image frames, the movement direction of the target object within that view is determined. For example, if the center coordinates of the later frame shift from the left to the right relative to the previous frame in the horizontal direction, the movement direction is determined to be horizontal to the right; if the center coordinates shift from the lower to the upper vertical direction, the movement direction is determined to be vertically upward. If both horizontal and vertical shifts exist simultaneously, they are combined to form a comprehensive description of the movement direction. During the continuous movement of the target object within the same monitoring field of view, the interval duration and offset amplitude of positional changes between adjacent image frames are statistically analyzed. When the interval duration is uniform and the offset amplitude remains stable, the movement rhythm is determined to be uniform speed. When the offset amplitude gradually increases while the interval duration remains unchanged, the movement rhythm is determined to be acceleration. When the offset amplitude gradually decreases while the interval duration remains unchanged, the movement rhythm is determined to be deceleration. If neither the offset amplitude nor the interval duration has a fixed pattern, the overall trend is combined to summarize a specific rhythmic characteristic. The movement direction and rhythmic trend of the same target object under all different monitoring fields are summarized, and the movement characteristics corresponding to each field of view are associated and bound to the information of that field of view, forming a complete record of the initial target information of the target object's cross-field of view movement state of the movement characteristics.

[0044] The motion consistency of the initial target information is obtained by evaluating the degree of fit of motion features under different monitoring fields of view based on multi-view geometric constraints. First, spatial association rules related to the target object to be evaluated are extracted from the multi-view geometric constraints, including the reasonable path range, imaging ratio relationship, and position mapping rules between different monitoring devices involved in the target object. The motion direction under different monitoring fields of view is extracted from the motion features of the initial target information. This motion direction is compared to see if it matches the reasonable path range for the target to move between corresponding devices as specified in the constraints. For example, if the constraints specify that the reasonable path for the target to move from the field of view of device A to the field of view of device B is horizontal forward, then it is determined whether the motion direction under both fields of view conforms to this horizontal forward path direction, without any reverse or deviation from the path. Simultaneously, the motion rhythm trend under different monitoring fields of view is compared, and combined with the imaging ratio relationship of different fields of view in the constraints, it is determined whether the motion rhythm remains consistent. For example, if a target is moving at a constant speed in field of view A, and the constraints specify that the imaging ratio does not change significantly in field of view B, then its motion rhythm should still remain constant, without sudden acceleration or deceleration, and should match the rhythm of field of view A. In cases of overlapping fields of view, the position mapping rules in the constraints are used to verify whether the movement direction and rhythm of the target object within the overlapping area are consistent with the movement characteristics in both fields of view, without any contradictions or conflicts. Based on the combined evaluation results of direction matching, rhythm coherence, and consistency of the overlapping area, if all evaluation dimensions meet the constraints, the degree of motion feature matching is determined to be high. If some dimensions meet the requirements but there are no core conflicts, the degree of matching is determined according to the core rules of the constraints, ultimately forming an initial target information motion consistency that accurately reflects the motion feature matching situation under different monitoring fields of view.

[0045] To establish preliminary cross-view trajectory association in underground space, target motion trajectory segments that meet preset conditions for motion consistency in the initial target information are correlated and matched. The preset conditions are defined as the core matching requirements where motion consistency is fully matched or meets the constraints, namely, the motion direction and reasonable path are completely consistent under different monitoring views, the motion rhythm and trend are coherent and conflict-free, and the motion characteristics in overlapping areas are completely corresponding. All target motion trajectory segments are extracted from the initial target information. Each trajectory segment contains the motion characteristics of the same target object under a single monitoring view and its corresponding basic attribute information. Based on the motion consistency evaluation results, combinations of trajectory segments that meet the preset conditions for motion consistency are selected. The selected trajectory segment combinations are further verified by comparing the morphological characteristics of the target object corresponding to each trajectory segment in the combination to ensure that the morphological characteristics are consistent and without significant differences. Simultaneously, the occurrence time of each trajectory segment is checked to ensure that the end time of the trajectory segment in the previous view and the start time of the trajectory segment in the next view are connected, conforming to the temporal logic of the target object's movement. Following chronological order and spatial movement paths, verified trajectory segments are sequentially connected. For example, the trajectory segment with the earliest ending time in view A is used as the starting segment, followed by trajectory segments in view B with consecutive starting times and consistent motion. This process is repeated for other qualified trajectory segments in subsequent views. During the connection process, the monitoring view information and motion characteristics corresponding to each trajectory segment are recorded, forming a complete cross-view trajectory chain. Each trajectory chain corresponds to the continuous movement of the same target object. After all trajectory chains are aggregated, a preliminary cross-view trajectory association for the underground space is formed, fully presenting the continuous movement trajectory of the target object across different monitoring views.

[0046] Numerical value indicating the degree of compatibility The result is derived from calculations of relevant monitoring data using this formula. This result is the product of calculations on the correlation of target information under different monitoring views. The number of samples within the preset time window is determined based on the actual needs of underground space safety monitoring. It represents the number of target object monitoring samples included within a pre-defined specific time range. These samples are entries selected from the continuous monitoring records of the target objects within the corresponding time window. The multi-view geometric constraints include the viewpoint... To view The fundamental matrix is ​​derived from the multi-view geometric constraints established after determining the relative positions of multi-view monitoring equipment based on the actual structural layout of the underground space, specifically addressing the perspective. Perspective The spatial location association rules between these two monitoring perspectives are a matrix constructed based on information such as the relative position coordinates of the devices and the field of view coverage characteristics. (Time) Time in perspective The target position coordinates are derived from the motion features of the initial target information, extracted over time. The target object in the view The location information corresponding to the surveillance image is used to determine the viewpoint at that moment. The coordinate information obtained is the center position of the pixel region of the target object in the image. Time Time in perspective The target position coordinates are derived from the motion features of the initial target information, extracted over time. The target object in the view The location information corresponding to the surveillance image is used to determine the viewpoint at that moment. The coordinate information obtained is the center position of the pixel region of the target object in the image. The preset scaling parameter used to normalize the epipolar error is a fixed value pre-set based on the imaging parameters and field-of-view coverage of the underground space monitoring equipment, used to adjust the numerical range of the epipolar error to suit the calculation requirements of the fit. The exponential function of the natural constant, derived from a mathematically defined function form, is used to transform the calculation results within the parentheses in the formula, ensuring that the final value meets the requirements of the fit.

[0047] The significance of this formula is to calculate the degree of fit. The specific process involves combining sample data within a preset time window, using the fundamental matrix in the multi-view geometric constraints to calculate the correlation between the target position coordinates at the same time under different views, and then processing the calculation result through the exponential function of the natural constant. The resulting value is the degree of fit, which reflects the degree of matching of the target motion characteristics under different monitoring views. The higher the degree of fit, the more the target motion characteristics under different monitoring views meet the requirements of the multi-view geometric constraints.

[0048] When the correlation results of the target position coordinates at the same time from different perspectives better meet the requirements of the multi-view geometric constraints, the deviation of the relevant numerical values ​​used in the formula will be smaller. In this case, the calculation result within the parentheses of the formula will be smaller, and after processing with the exponential function of the natural constant, the value of the fit will be larger. Conversely, when the deviation between the correlation results of the target position coordinates at the same time from the requirements of the multi-view geometric constraints is greater, the deviation of the relevant numerical values ​​used in the formula will be greater. In this case, the calculation result within the parentheses of the formula will be larger, and after processing with the exponential function of the natural constant, the value of the fit will be smaller. When the overall deviation of the relevant numerical values ​​corresponding to all samples within the preset time window is small, the value of the fit calculated by the formula will be at a high level. Conversely, when the overall deviation of the relevant numerical values ​​of these samples is large, the value of the fit will be at a low level.

[0049] In a 200-meter-long tunnel test section, two high-definition cameras with overlapping fields of view were deployed: Cam_A was installed at K0+050, facing due north, with a field of view of 70°; Cam_B was installed at K0+100, also facing north, with a field of view of 65°. The overlapping area of ​​the two cameras was approximately from K0+080 to K0+120.

[0050] First, establish multi-view geometric constraints: through a prior calibration process, use Zhang Zhengyou's calibration method to obtain the internal parameter matrix and external parameters of the two cameras, then calculate the fundamental matrix from Cam_A to Cam_B, and combine it with the tunnel design drawings to determine that the reasonable movement path of the target between the two cameras should be along the tunnel axis.

[0051] Next, motion features of the same engineering vehicle in Cam_A and Cam_B were extracted from the initial target information: In the view of Cam_A, the vehicle's pixel coordinates moved from (200, 300) to (280, 300) within 5 consecutive frames, and the calculated direction of motion was horizontal to the right, with a speed of approximately 1.5 m / s; in the view of Cam_B, the target pixel coordinates changed from (150, 320) to (230, 320) within the same time period, and the direction of motion was also horizontal to the right. Using the fit calculation formula for evaluation, since the fit was higher than the preset threshold of 0.85, the system determined that the trajectory segments in the two views had a high degree of motion consistency. Finally, the trajectory segment of the vehicle in Cam_A from 14:30:22 to 14:30:23 was associated and matched with the segment in Cam_B from 14:30:23 to 14:30:24, forming a preliminary continuous trajectory across viewpoints, with a recorded association confidence of 0.92.

[0052] The beneficial effects include establishing multi-view geometric constraints by combining the actual structural layout of underground space, accurately defining the relative positions, field-of-view correlations, and reasonable rules for target movement of multi-view monitoring equipment, providing a solid spatial basis for evaluating motion characteristics under different monitoring views, and avoiding the blindness of cross-view analysis. It accurately identifies the movement direction and rhythm trend of the same target object under different monitoring views in the initial target information, fully extracts the motion characteristics reflecting the target's motion state, comprehensively captures the core attributes of the target's cross-view movement, and lays a detailed data foundation for subsequent motion consistency assessment. Evaluating the degree of fit of motion characteristics based on multi-view geometric constraints can accurately determine the matching of target motion states under different monitoring views, effectively eliminate invalid data with contradictory motion characteristics, and ensure the accuracy and reliability of motion consistency determination. By associating and matching target motion trajectory segments that meet preset conditions for motion consistency, a preliminary cross-view trajectory association for underground space was successfully established. This fully presented the continuous motion process of the target object between different monitoring views, breaking the monitoring limitations of a single view. It significantly improved the integrity and coherence of the target motion trajectory, providing high-quality basic data support for further optimization of cross-view trajectories and accurate understanding of the dynamic target movement patterns in underground space, thus enhancing the comprehensiveness and effectiveness of underground space safety monitoring.

[0053] S4. Based on the preliminary cross-view trajectory association, the initial target information is fused into a continuous trajectory under a unified coordinate system to generate the fused trajectory data of the underground space; In this embodiment of the invention, the step of fusing the initial target information into a continuous trajectory in a unified coordinate system based on the preliminary cross-view trajectory association to generate the fused trajectory data of the underground space includes: A unified spatial coordinate system for the underground space is established by using a pre-set fixed reference marker within the underground space as a reference. The trajectory segments in the preliminary cross-view trajectory association are mapped to the unified coordinate system to calibrate the spatial position of the trajectory segments, thereby obtaining the calibrated trajectory segments of the underground space; Determine the rationality of the connection between the trajectory segments in order to retain the trajectory connection relationships with high consistency among the trajectory segments; Based on the trajectory connection relationship, reasonable transition information at the connection point of the calibration trajectory segment is supplemented to obtain the fused trajectory data of the underground space.

[0054] A unified spatial coordinate system is established using pre-set fixed reference markers within the underground space. First, these fixed reference markers are identified; they are immovable marks pre-installed on the permanent structure of the underground space, including special markings on wall surfaces, fixed marks on column sides, and embedded marks on the ground. The positions of these markers are fixed during the underground space construction phase and will not shift. One fixed reference marker located in the core area of ​​the underground space is selected as the origin of the unified spatial coordinate system. The extension direction of the main underground passage is taken as the primary horizontal axis, the horizontal direction perpendicular to the main passage is taken as the secondary horizontal axis, and the direction perpendicular to the ground is taken as the vertical axis. The correspondence between the extension directions of these three axes and the actual structure of the underground space is clarified. Measuring tools are used to determine the positions of all other fixed reference markers in this coordinate system one by one, recording the positional information of each marker along its corresponding axis. By comparing the relative positions of different markers, the consistency between the spatial scale of the coordinate system and the actual physical scale of the underground space is verified, ensuring that the coordinate system accurately maps all areas of the underground space without scale deviation or positional misalignment. Ultimately, a unified spatial coordinate system covering the entire underground space was established. This coordinate system can uniquely associate any physical point in the underground space with its corresponding coordinate information, providing a unified positional reference system for the spatial calibration of subsequent trajectory segments.

[0055] The trajectory segments from the initial cross-view trajectory association are mapped to a unified coordinate system to calibrate their spatial positions, resulting in calibrated trajectory segments. All trajectory segments are extracted from the initial cross-view trajectory association, each containing the target object's motion position information within a single monitoring field of view. The deployment information of the monitoring equipment corresponding to each trajectory segment is queried to obtain the equipment's installation position coordinates in the unified spatial coordinate system, as well as parameters such as the lens's installation angle and field of view, clarifying the transformation relationship between the equipment's local field of view coordinates and the unified spatial coordinate system. For each location point in a trajectory segment, based on the corresponding monitoring equipment parameters, the center position of the pixel area within the equipment's local field of view is converted to its actual physical location in the unified spatial coordinate system. During the conversion, the directional deviation of the location is adjusted according to the equipment's shooting angle, and the scale ratio of the location is adjusted according to the equipment's effective monitoring distance to ensure that the converted location accurately corresponds to the actual area within the underground space. After completing the conversion of all location points for each trajectory segment, the converted location points are rearranged in chronological order. The relative distances between the location points are checked to ensure they match the target object's motion characteristics; for example, the interval between location points corresponding to uniform motion should remain stable. If deviations exist, the conversion parameters are readjusted until a match is achieved. Finally, calibration trajectory segments of the underground space are obtained. All position points of each calibration trajectory segment correspond to the actual physical position in a unified spatial coordinate system, ensuring that the position information of different trajectory segments are under the same spatial reference system.

[0056] The rationality of the connection between trajectory segments is determined to retain highly consistent trajectory connections. The criteria for determining the rationality of trajectory segment connections are clarified, including whether the distance between the end and start positions of two trajectory segments in a unified spatial coordinate system is within a reasonable range for a single movement of the target object; whether the movement directions of the two trajectory segments are continuous without significant abrupt changes; and whether the movement rhythms of the two trajectory segments are consistent without significant differences. All trajectory connection relationships are extracted from the initial cross-view trajectory association, with each connection relationship corresponding to two consecutive trajectory segments. For each connection relationship, the straight-line distance between the last position of the previous trajectory segment and the first position of the next trajectory segment in a unified spatial coordinate system is first checked. If this distance does not exceed the maximum distance the target object can move within the corresponding time interval, the distance dimension is deemed to meet the rationality requirements. Then, the ending movement direction of the previous trajectory segment is compared with the starting movement direction of the next trajectory segment. If the deviation between the two directions is within the normal turning angle range of the target object, the direction dimension is deemed to meet the rationality requirements. Finally, the movement rhythms of the two trajectory segments are checked. If the rhythm of the first segment is uniform and the rhythm of the second segment is also uniform, or if the acceleration rhythm of the first segment is continuous with the acceleration rhythm of the second segment, then the rhythm dimension is deemed to meet the rationality requirements. Only when all three dimensions—distance, direction, and rhythm—meet the rationality requirements are the trajectory connection relationships considered to have a high degree of fit and are retained. Connection relationships that do not meet the requirements are directly excluded, ultimately resulting in the retained trajectory connection relationships with a high degree of fit.

[0057] Based on the trajectory connection relationships, reasonable transition information at the connection points of the calibrated trajectory segments is supplemented to obtain the fused trajectory data of the underground space. For each retained trajectory connection relationship, two corresponding calibrated trajectory segments are extracted, and the last position point of the previous trajectory segment is determined as the connection start point, and the first position point of the subsequent trajectory segment is determined as the connection end point. Based on the motion characteristics of the target object in these two trajectory segments, such as a horizontal forward motion direction and a constant speed, a reasonable movement path between the connection start point and the connection end point is calculated. This path must conform to the actual structural layout of the underground space, such as following the extension direction of the passage rather than penetrating the wall. According to the motion rhythm of the target object, several transition position points are added between the connection start point and the connection end point. The time interval of each transition position point is consistent with the time interval of the original trajectory segment, and the position interval is determined according to the motion rhythm. For example, the position interval corresponding to constant speed motion is the same as the position interval of the original trajectory segment. The supplemented transition position points are inserted between the two trajectory segments in chronological order, so that the connection start point, transition position points, and connection end point form a continuous position sequence. After supplementing transition information for all trajectory segments corresponding to the retained connection relationships, all associated calibration trajectory segments and transition point are integrated into a continuous trajectory sequence. The positional coherence, directional consistency, and rhythmic stability of the entire sequence are checked to ensure there are no interruptions or contradictions. Finally, fused trajectory data of the underground space is obtained, which presents the continuous movement trajectory of the target object in the underground space and can completely reflect the movement process of the target object.

[0058] A local independent coordinate system for the tunnel is established with a permanent surveying control point (numbered CP01, whose three-dimensional coordinates in the National Geodetic Coordinate System are X=1000.000m, Y=2000.000m, Z=50.000m) at the tunnel entrance as the origin: the positive direction of the X-axis is along the tunnel centerline pointing into the tunnel, the Y-axis is perpendicular to the tunnel centerline and horizontally to the left, and the Z-axis is vertically upward.

[0059] Map the aforementioned cross-view trajectory segments to this coordinate system: for trajectory points in Cam_A, back-project using camera calibration parameters, and combine with pixel depth information to transform the image coordinates to the world coordinate system; process trajectory points in Cam_B in the same way.

[0060] For example, a pixel (250, 300) in Cam_A is transformed to (1005.2, 1998.7, 50.3) in a unified coordinate system, while the corresponding pixel (180, 310) in Cam_B is transformed to (1005.8, 1998.5, 50.2). Then, a reasonable connection is determined: the Euclidean distance between the last coordinate point (1010.1, 1999.0, 50.5) of Cam_A and the first coordinate point (1010.3, 1998.9, 50.5) of Cam_B is calculated to be 0.22 meters, which is less than the maximum possible distance the vehicle could move within a 1-second time interval; the direction of movement at the connection point for both segments is along the positive X-axis; and the movement rhythm is uniform.

[0061] Therefore, the connection was deemed reasonable. The system then performed B-spline interpolation between the two points to add two transition points, such as (1010.15, 1998.95, 50.5) and (1010.25, 1998.92, 50.5), ultimately generating a time-continuous and spatially smooth fused trajectory in a unified coordinate system. This trajectory contains 15 three-dimensional coordinate points, has a total duration of 3 seconds, and describes the complete movement process of the vehicle from K0+060 to K0+090.

[0062] The beneficial effects include establishing a unified spatial coordinate system for the underground space through fixed benchmarks, constructing a consistent positional reference system covering the entire underground space, eliminating spatial scale deviations and positional misalignments, and providing an accurate and unified reference for subsequent spatial calibration of trajectory segments. Mapping trajectory segments from the initial cross-view trajectory association to the unified coordinate system yields calibrated trajectory segments. By transforming the positions within the local field of view of the monitoring equipment to the unified coordinate system, and adjusting direction and scale deviations based on equipment parameters, while simultaneously matching the target motion characteristics to calibrate the positions, the positional information of different trajectory segments is placed within the same spatial system, improving the spatial accuracy of the trajectory segments. The rationality of the connection between trajectory segments is determined, and highly consistent trajectory connections are retained. The rationality of the connection is verified from three dimensions: distance, direction, and rhythm, eliminating unsuitable connections, ensuring the reliability and consistency of trajectory connections, and avoiding interference from unreasonable connections to the trajectory data. By supplementing and calibrating the reasonable transition information at the junctions of trajectory segments according to the trajectory connection relationship, the fused trajectory data is obtained. By supplementing the transition position points according to the underground space structure and target movement characteristics, the data is integrated to form a continuous trajectory sequence, which fully presents the movement process of the target object in the underground space. This improves the coherence and integrity of the trajectory data and provides high-quality continuous trajectory evidence for the monitoring and analysis of dynamic targets in underground space.

[0063] S5. Project the fused trajectory data into a three-dimensional space, and perform a natural object fall source tracing analysis on the fused trajectory data to obtain the source location information of the underground space; In this embodiment of the invention, the step of projecting the fused trajectory data into a three-dimensional space and performing a natural object fall source tracing analysis on the fused trajectory data to obtain the source location information of the underground space includes: Based on the actual three-dimensional structural data of the underground space, determine the correspondence rules between the fused trajectory data and the physical location in the three-dimensional space; Based on the correspondence, the fused trajectory data is projected onto the three-dimensional space to obtain the three-dimensional trajectory data of the underground space; By filtering the three-dimensional trajectory data to obtain the trajectory fragments of the natural objects, the trajectory data corresponding to the natural objects are obtained. Based on the trajectory supplementary information recorded by the multi-view monitoring equipment in the underground space, the trajectory deviation of the natural object movement trajectory segment is corrected; By tracing the starting point of the corrected trajectory of the natural object and combining it with the position markers in the three-dimensional space, the starting spatial coordinates of the natural object's fall are determined. Based on the specific regional division of the underground space, the spatial positioning verification of the starting spatial coordinates is performed to obtain the source location information of the underground space.

[0064] Based on the actual 3D structural data of the underground space, the correspondence rules between the fused trajectory data and the physical locations in 3D space are determined. First, the actual 3D structural data of the underground space is acquired. This data covers the 3D shape, dimensions, and relative positions of the walls, columns, passageways, tops, and bottoms of the underground space, including the horizontal extension range and vertical height range of each structure, as well as their interconnections. The coordinate information of each location point in the fused trajectory data is analyzed. Combined with the physical spatial characteristics in the 3D structural data, the correspondence between the horizontal coordinates of the fused trajectory data and the horizontal and vertical regions of the passageways in 3D space is clarified. That is, the horizontal coordinate values ​​of the fused trajectory correspond to the actual horizontal and vertical positions of the passageways in the 3D structure, and the vertical coordinates of the fused trajectory correspond to the height range from top to bottom in 3D space. Simultaneously, considering the actual size ratio of the 3D structure, the mapping relationship between the fused trajectory coordinates and the physical locations is adjusted to ensure that each location point in the fused trajectory data corresponds to a unique physical location in 3D space, without overlap or misalignment. Finally, the correspondence rules between the fused trajectory data and the physical locations in 3D space are determined. These rules clarify the mapping method from the fused trajectory coordinates to the 3D physical locations, providing a clear basis for subsequent projection operations.

[0065] Based on the correspondence, the fused trajectory data is projected into 3D space to obtain 3D trajectory data for the underground space. First, the established correspondence rules are retrieved, and the fused trajectory data is broken down into individual location points in chronological order. For each location point, according to the mapping method in the correspondence rules, the horizontal coordinates of the fused trajectory data are converted into corresponding horizontal and vertical physical positions in 3D space, and the vertical coordinates are converted into corresponding height physical positions in 3D space. Simultaneously, structural features from the 3D structural data are considered to ensure that the converted positions are within a reasonable area in 3D space, such as avoiding penetration through walls. After all location points have been converted, these 3D physical positions are connected sequentially in chronological order to form continuous trajectory lines, with each line node corresponding to a time point in the fused trajectory data. The connected trajectories are checked to confirm that the trajectory lines match the physical space of the 3D structural data, with no positional deviations or logical contradictions. Finally, the 3D trajectory data for the underground space is obtained, which fully presents the physical path of the fused trajectory in 3D space.

[0066] The process involves filtering trajectory data corresponding to natural objects from 3D trajectory data to obtain motion trajectory fragments of the natural objects. First, the trajectory characteristics corresponding to the natural objects are identified. The motion trajectories of natural objects typically exhibit a vertically downward or diagonally downward direction, with a continuous acceleration rhythm and no active turning. Furthermore, the starting position of the trajectory is often located in the top structure area or around pipelines in underground spaces. Each trajectory in the 3D trajectory data is traversed, and the motion direction of each trajectory is analyzed to determine if it conforms to the vertical or diagonally downward characteristics. The motion rhythm of the trajectory is then analyzed to confirm whether it is in a continuous acceleration state. Simultaneously, it is checked whether the starting position of the trajectory is in the top structure or around pipelines. Trajectories that simultaneously meet these characteristics are extracted, and the position points in the trajectory are arranged in chronological order to ensure the continuity of the trajectory and consistency of time. Trajectory data that does not conform to the trajectory characteristics of natural objects are excluded. Finally, the motion trajectory fragments of the natural objects are obtained, which only contain the 3D trajectory information corresponding to the natural objects.

[0067] The trajectory deviation of natural object movement segments is corrected based on supplementary trajectory information recorded by multi-view monitoring equipment in underground space. First, supplementary trajectory information recorded by the multi-view monitoring equipment is collected. This information includes the actual position details of the natural object at each time point under different monitoring views, as well as the relative positional relationship with surrounding structures. Each position point in the natural object's movement trajectory segment is compared with the corresponding position in the supplementary information to identify deviation points. For each deviation point, the three-dimensional coordinates of the point are adjusted according to the actual position in the supplementary information. For example, if the supplementary information shows that the natural object is located next to a support in the top structure at a certain time point, but the position of this point in the trajectory segment is off-center, the coordinates of the point are adjusted to the corresponding position next to the support, while ensuring that the adjusted point maintains continuity with the movement trajectory of the preceding and following positions. After adjusting all deviation points, the trajectory segment and supplementary information are compared again to confirm that no deviation exists. Finally, the corrected natural object movement trajectory segment is obtained, whose position information perfectly matches the monitoring supplementary information.

[0068] The starting point of the corrected trajectory of the natural object is traced back, and the initial spatial coordinates of its fall are determined by combining this with location markers in three-dimensional space. From the corrected trajectory fragments, the earliest recorded location point is located in chronological order; this point is the starting point of the trajectory. Structural data in three-dimensional space is retrieved, and location markers around this starting point are examined. These markers include structural components with clearly defined physical locations, such as fixed parts of the top structure and support nodes for pipelines. By comparing the relative positions of the starting point and these location markers, the specific physical location of the starting point in three-dimensional space is determined. For example, if the starting point is located in the right region of a top support, the three-dimensional coordinates of the starting point are determined by combining the three-dimensional coordinates of the support and its relative position. The determined coordinates are verified to confirm that they fall within the corresponding physical structural area in three-dimensional space without any positional errors. Finally, the initial spatial coordinates of the natural object's fall are obtained, clearly identifying the three-dimensional physical location where the fall began.

[0069] Based on the specific zoning of underground space, spatial positioning verification of the initial spatial coordinates is performed to obtain the source location information of the underground space. First, the specific zoning of the underground space is clearly defined. These zonings are bounded by fixed structures, including the top structure area, the area above passageways, and the pipeline layout area. The boundary of each area is determined by the position of fixed structures such as walls and columns. The physical location corresponding to the initial spatial coordinates is determined, and it is verified whether the coordinates are within the boundary of that area. For example, if the location corresponding to the initial spatial coordinates is within the boundary of the top structure area, it is confirmed that the coordinates belong to the top structure area. Simultaneously, combined with specific structural information within that area, such as the distribution of supports, the specific sub-location of the initial spatial coordinates within that area is further clarified. The initial spatial coordinates are combined with the corresponding specific sub-location information of the area to form complete positioning content. This content is then verified a second time to confirm the matching accuracy between the area and the coordinates, ultimately obtaining the source location information of the underground space. This information clearly identifies the area and specific location corresponding to the starting point of the natural object fall.

[0070] First, actual 3D structural data is obtained from the tunnel's BIM model, including the precise curved surface model of the tunnel lining segments, the geometry of the track bed, and the spatial location of the overhead suspended equipment. The correspondence between the fused trajectory data and 3D space is determined: the X-coordinate in a unified coordinate system corresponds to the "mileage along the tunnel" in the BIM model, the Y-coordinate corresponds to the "lateral offset," and the Z-coordinate corresponds to the "height above the track surface," with scales standardized. The fused trajectories of the aforementioned engineering vehicles are imported into the BIM software and superimposed onto the 3D tunnel model via a coordinate transformation interface, generating corresponding 3D spatial polylines. Subsequently, a falling object detection algorithm is run: trajectory segments with continuously increasing vertical velocity and horizontal velocity less than 0.5 m / s are selected from all 3D trajectories. For example, the system selects a trajectory segment whose Z-coordinate decreases from 8.6 meters to 5.2 meters within 0.5 seconds, calculating an average vertical velocity of 6.8 m / s, which meets the characteristics of free fall and is determined to be a "natural object falling."

[0071] Next, the system uses supplementary trajectory information from multi-view monitoring equipment for correction. For example, another side-view camera, Cam_C, shows that the falling object was closer to the tunnel sidewall at the initial moment, so the Y coordinate of the starting point of the 3D trajectory is slightly adjusted from 1999.0 to 1998.3. After correction, the system traces back along the trajectory point to determine its starting endpoint's 3D coordinates as (1120.5, 1998.3, 8.6). Combined with the location marker in the BIM model, this point is located at "section tunnel K11+120.5, top of the 35th ring segment, adjacent to the lighting fixture mounting bracket".

[0072] Finally, based on the tunnel's operation and maintenance management area division, it was confirmed that the coordinates belonged to "11th Management Section, Top Equipment Area". The coordinates were then compared and verified with the "Light Fixture Bracket Bolt Embedded Point" location in the inspection record. The deviation was less than 0.1 meters, thus outputting accurate source location information: "The falling object started at K11+120.5, the top of the 35th ring segment, 0.3 meters west of the light fixture bracket".

[0073] The beneficial effects include: determining the correspondence between fused trajectory data and 3D physical location by combining actual 3D structural data of underground space; clarifying the precise mapping method between trajectory coordinates and 3D space; providing a scientific basis for trajectory projection; and ensuring that subsequent 3D trajectory data can accurately reflect the physical spatial path. Based on the correspondence, the fused trajectory data is projected into 3D space to obtain 3D trajectory data, realizing the transformation of trajectory from planar to 3D space, fully presenting the three-dimensional movement path of the target object in underground space, making the trajectory information more consistent with the actual spatial scene. Movement trajectory segments corresponding to natural objects are filtered from the 3D trajectory data. Precise filtering is achieved by clarifying the unique characteristics of natural object trajectories, eliminating interference from non-natural object trajectories, focusing on core monitoring targets, and laying the foundation for tracing the source of natural object falls. Deviations in natural object movement trajectory segments are corrected based on supplementary trajectory information from multi-view monitoring equipment. Trajectory coordinates are adjusted by comparing actual location details recorded by monitoring, eliminating positional errors in the trajectory data and improving the accuracy and reliability of natural object trajectories. The starting endpoint of the corrected trajectory is traced and combined with 3D spatial location markers to determine the initial spatial coordinates of the natural object's fall, accurately locking the initial physical location of the fall behavior and providing core coordinate basis for tracing the source. Based on the division of underground space areas, the starting spatial coordinates are located and verified to obtain the source location information. The coordinates are associated with specific areas and sub-locations to ensure the accuracy and clarity of the source location. This fully realizes the tracing of the starting location of natural object falls, providing accurate location references for underground space safety risk investigation and hidden danger management, and enhancing the pertinence and effectiveness of underground space safety monitoring.

[0074] S6. Integrate the source location information with the preset safety knowledge graph to generate visual data on the safety situation of the underground space.

[0075] In this embodiment of the invention, the step of integrating the source location information with a preset security knowledge graph to generate visualized security situation data of the underground space includes: Based on the spatial identifier in the source location information, match the corresponding spatial region and associated security rules in the preset security knowledge graph; Based on the safety rules, the types of natural objects corresponding to the source location information are associated with the types of hazards in the safety knowledge graph to determine the potential safety risk attributes of the underground space; Verify the logical consistency between the source location information and the potential security risk attributes to correct the source location information; Multimodal evolution is performed on the core security information in the corrected source location information to obtain the security situation visualization data of the underground space.

[0076] The process of performing multimodal evolution on the core security information in the corrected source location information to obtain the security situation visualization data of the underground space includes: According to the preset multimodal presentation specifications, the text-based association information of the core security elements in the corrected source location information is converted into standardized explanatory text to obtain the standardized text data of the corrected source location information. Using preset graphical identification rules, the spatial association data in the core safety elements is graphically transformed to obtain the spatial graphical data of the underground space; Add time-series association markers to the standardized text data and the spatial graphical data to bind the state change trend of core security information with the time dimension, thereby obtaining the multimodal integrated data of the underground space; Based on the requirements of the underground space security situation display scenario, duplicate information elements in the multimodal integrated data are removed, and the information complementarity and logical coherence between different modal data are strengthened to obtain the underground space security situation visualization data.

[0077] Based on the spatial identifiers in the source location information, a matching process is performed with the corresponding spatial regions and associated safety rules in a pre-defined safety knowledge graph. First, complete spatial identifiers are extracted from the source location information, including the name of the area to which the location belongs and specific structural identifiers of the surrounding structures. Area names include, for example, the top structure area and pipeline layout area; specific structural identifiers include, for example, top support, pipeline nodes, and wall fasteners. The pre-defined safety knowledge graph is a pre-constructed structured information set containing all regional categories of underground space and their corresponding safety regulations. Each regional category in the graph is explicitly bound to unique spatial identifier features, and each category is associated with safety rules tailored to the safety requirements of that area. These rules cover prohibited hazard types, safety risks to be prevented, and protection standards against falling objects within the area. During matching, the extracted spatial identifiers are compared one by one with the spatial identifier features of each regional category in the safety knowledge graph. The regional category with a perfect match is identified; this category represents the spatial region corresponding to the source location information. Then, retrieve all the security rules bound to the spatial region from the map to ensure that each rule is highly compatible with the physical structure, function and security focus of the region. Finally, complete the accurate matching of spatial identifiers with spatial regions and security rules to obtain the corresponding spatial region and associated security rules.

[0078] The potential safety risk attributes of underground spaces are determined by associating the natural object types corresponding to the source location information with the hazard source types in the safety knowledge graph based on safety rules. First, the natural object types corresponding to the source location information are identified. These types are confirmed through previous trajectory feature analysis and supplementary monitoring information, such as stones, dust, pipeline debris, and wall spalling. Each safety rule in the safety knowledge graph corresponds to a specific range of hazard source types. These ranges are set according to the protection objectives of the rule. For example, for safety rules targeting the roof structure area, the corresponding hazard source type range includes falling objects and suspended particulate matter; for safety rules targeting pipeline areas, the corresponding hazard source type range includes pipeline appurtenance detachment and pipeline rupture fragments. During the association process, the core protection focus of the safety rule is first analyzed to clarify the range of hazard source types targeted by the rule. Then, the identified natural object types are precisely compared with this range to determine whether the natural object type belongs to the hazard source type corresponding to the rule. When a natural object type perfectly matches a certain hazard source type, the potential safety risk description associated with that hazard source type is extracted. For example, the potential safety risk attributes corresponding to falling object hazard sources are collision damage risk, passage blockage risk, etc., and the potential safety risk attributes corresponding to suspended particulate matter hazard sources are air pollution risk, equipment wear risk, etc., and finally the potential safety risk attributes of underground space are determined.

[0079] The logical consistency between the source location information and potential security risk attributes is verified to correct the source location information. The verification is based on the inherent correspondence between spatial regions and potential security risk attributes in a pre-defined security knowledge graph. Specifically, a specific spatial region can only be associated with a specific type of potential security risk attribute. If a spatial region and risk attribute do not match, it is considered a logical contradiction. During verification, the pre-defined risk attribute range corresponding to the spatial region of the source location information is first retrieved from the security knowledge graph. Then, the currently determined potential security risk attribute is compared with this range. If the potential security risk attribute is within the pre-defined range, the two are considered logically consistent, and the source location information does not need correction. If the potential security risk attribute exceeds the pre-defined range, indicating a logical contradiction, the source location information must be corrected. During the correction process, the trajectory supplementary information recorded by the multi-view monitoring equipment, the three-dimensional structural data of the underground space, and the movement trajectory fragments of natural objects were retrieved again. The regional division basis and specific location markers of the source location were checked again to check for any errors in regional judgment or location coordinate deviations. For example, if a location that originally belonged to a pipeline laying area was mistakenly identified as a passage area, the regional classification and coordinate labeling were readjusted until the spatial area corresponding to the corrected source location information could fully include the current potential safety risk attributes within its preset risk attribute range, ensuring that the two are logically consistent and conflict-free, and finally the corrected source location information was obtained.

[0080] Multimodal evolution of core safety information from the corrected source location information yields visualized safety situation data for underground space. This core safety information includes key elements such as the corrected source location area, specific location markers, natural object types, and potential safety risk attributes. Multimodal evolution transforms this textual core safety information into an intuitive and perceptible visual form. Specifically, based on the three-dimensional structural data of the underground space, a 3D visualization model with the same scale as the actual space is generated. Within the model, the starting position of the falling natural object is precisely marked using dedicated visualization icons according to the corrected source location coordinates. The icon style is set according to the type of natural object; for example, circular icons represent rocks, and square icons represent pipeline fragments. The icon color corresponds to the potential safety risk attribute, with red for high risk and yellow for medium risk, etc. A text description module is then added to the 3D model, clearly presenting the area name of the source location, specific structural markers, natural object types, and detailed descriptions of potential safety risk attributes. Simultaneously, a risk impact range visualization layer is generated. Based on the degree of impact of the potential safety risk attributes, color areas with different transparency are used to outline the spatial range that the risk may affect within the 3D model. Finally, by integrating the 3D model, labels, text descriptions, and risk impact range layers, a complete safety situation visualization data is formed. This data can intuitively and comprehensively display the source location information of underground spaces and the corresponding safety risk status, making it easy for staff to quickly grasp the core safety situation.

[0081] Based on the pre-defined multimodal presentation specifications, the textual association information of core security elements in the corrected source location information is converted into standardized explanatory text to obtain standardized text data of the corrected source location information. Textual association information of core security elements is extracted, covering areas such as region names, specific location identifiers, natural object types, and potential security risk attributes. Information is sorted according to the structural framework of the multimodal presentation specifications, and the text is optimized and redundancy removed based on expression logic. Standardized expression style and format are used according to terminology standards, integrating them to form a standardized explanatory text with a unified structure and standardized expression.

[0082] Using pre-defined graphical identification rules, spatial correlation data in core safety elements is graphically transformed to obtain spatial graphical data of the underground space. Information such as three-dimensional coordinates, area boundaries, risk impact range, and fall trajectory paths are analyzed from this spatial correlation data. Following the graphical identification rules, specific node icons are used to mark the source location, continuous lines are used to outline area boundaries, transparent filled areas represent the risk range, and arrowed line segments display the fall trajectory. All graphical elements are integrated to form complete spatial graphical data.

[0083] By adding time-series association markers to standardized text data and spatial graphical data, the state change trends of core safety information are bound to the time dimension to obtain multimodal integrated data of underground space. Key time points such as the start time of natural object falls and the time of risk attribute confirmation are extracted, and unified time-series association markers corresponding one-to-one with time points are generated. These markers are then bound to each information module of the standardized text data and each graphical element of the spatial graphical data, ensuring that the same marker is used for text and graphics at the same time point, and the data is arranged chronologically to form a complete time series.

[0084] Based on the requirements of underground space security situation display scenarios, redundant information elements in multimodal integrated data are removed, and the information complementarity and logical coherence between different modalities are enhanced to obtain underground space security situation visualization data. Duplicate information is identified according to the display scenario requirements, and only one presentation format is retained. Text data supplements details not intuitively presented by graphics, and graphic data supplements the abstract spatial relationships in text. The logical order of the data is sorted out, and the integrated and optimized text and graphic data form security situation visualization data adapted to the scenario requirements.

[0085] The system's built-in safety knowledge graph is constructed using the Neo4j graph database and includes nodes such as "spatial area," "hazard source type," and "safety rule" and their relationships. Based on the spatial identifier "K11+120.5, top of the 35th ring segment" in the source location information, the corresponding spatial area node in the graph is matched as "tunnel top equipment installation area." This node is associated with three safety rules: "Rule ID: TD-01: The top area needs to prevent the falling of metal parts such as bolts and brackets," "Rule ID: TD-02: The radius of influence of falling objects is within 5 meters below the starting point," and "Rule ID: TD-03: The connection security of nearby equipment needs to be checked."

[0086] Based on rule TD-01, the system associates the traced natural object type "metal bolt" with the hazard source type node "falling object - metal part" in the knowledge graph, thereby determining the potential safety risk attributes as "mechanical impact risk" and "equipment damage risk".

[0087] Logical consistency verification was performed: The risk attributes preset in the "tunnel top equipment installation area" of the knowledge graph include "mechanical impact" and "electrical short circuit". The currently identified risk attributes are within its range, and the verification passed.

[0088] Subsequently, multimodal evolution was performed: In terms of text, standardized explanatory text was generated based on the specification template: "Alarm time: 2023-10-10 14:35:10; Event type: Object falling; Location: Top of the 35th ring at K11+120.5; Object description: Metal bolt; Risk level: Medium; Impact range: 5 meters below; Handling suggestion: Immediately isolate the area and inspect adjacent supports." Graphically, in the tunnel's 3D BIM model, a dynamically flashing red cone icon was overlaid at coordinates (1120.5, 1998.3, 8.6), and a semi-transparent red cone was generated downwards from this point to represent the risk impact area.

[0089] Meanwhile, this location is marked with a red exclamation mark on the two-dimensional planar diagram. In terms of timing, all visualization elements are linked to the timestamp "14:35:10," and the scene before and after the fall can be replayed on the timeline. Ultimately, the system integrates text reports, 3D situation maps, 2D planar diagrams, and timeline controls to generate an interactive security situation visualization interface for real-time decision-making by monitoring center personnel.

[0090] The beneficial effects include: accurately establishing the correspondence between location information and safety regulations by matching spatial identifiers in the source location information with corresponding spatial regions and associated safety rules in a pre-defined safety knowledge graph; avoiding the blind application of safety rules; and providing a rule-based basis for subsequent risk assessment that fits the actual scenario. Based on the association of safety rules with natural object types and hazard source types in the knowledge graph, precise binding of natural objects with potential risks is achieved, clarifying the potential safety risk attributes of underground spaces, making risk identification more targeted, and effectively avoiding biases in risk assessment. Verifying the logical consistency between source location information and potential safety risk attributes and correcting the source location information involves comparing the inherent correspondence between spatial regions and risk attributes in the knowledge graph, identifying and correcting errors in location information judgment, ensuring logical consistency between source location and risk attributes, and improving the accuracy and reliability of source information. Multimodal evolution of core safety information in the corrected source location information yields visualized safety situation data. This transforms text-based core safety information into an intuitive visualization format that includes 3D models, unique icons, text descriptions, and risk range layers. This breaks down the barriers to understanding textual information, enabling staff to quickly and comprehensively grasp the source location of underground spaces and the corresponding safety risk status. It significantly improves the efficiency of safety situation perception and provides clear and intuitive data support for the timely investigation and precise management of safety hazards in underground spaces, thereby enhancing the practicality and effectiveness of safety monitoring.

[0091] like Figure 2 The diagram shown is a functional module diagram of an underground space safety situation edge data visualization system provided in an embodiment of the present invention.

[0092] The underground space safety situation edge data visualization system 100 described in this invention can be installed in an electronic device. Depending on the functions implemented, the underground space safety situation edge data visualization system 100 may include a data acquisition module 101, a target detection module 102, a trajectory association module 103, a trajectory fusion module 104, a source tracing analysis module 105, and a visualization generation module 106. The modules described in this invention can also be referred to as units, which are a series of computer program segments that can be executed by the processor of an electronic device and can perform a fixed function, and are stored in the memory of the electronic device.

[0093] In this embodiment, the functions of each module / unit are as follows: The data acquisition module 101 is used to acquire standardized data of underground space; The target detection module 102 is used to perform dynamic target detection on the standardized data to obtain the initial target information of the underground space; The trajectory association module 103 is used to evaluate the motion consistency of the initial target information under different views based on the multi-view geometric constraints of the underground space, so as to establish a preliminary cross-view trajectory association of the underground space. The trajectory fusion module 104 is used to fuse the initial target information into a continuous trajectory in a unified coordinate system based on the preliminary cross-view trajectory association, so as to generate the fused trajectory data of the underground space. The source tracing analysis module 105 is used to project the fused trajectory data into a three-dimensional space and perform natural object fall source tracing analysis on the fused trajectory data to obtain the source location information of the underground space. The visualization generation module 106 is used to integrate the source location information with the preset safety knowledge graph to generate the safety situation visualization data of the underground space.

[0094] In the several embodiments provided by this invention, it should be understood that the disclosed methods and systems can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.

[0095] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0096] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0097] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0098] This application embodiment can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0099] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for visualizing edge data on the safety situation of underground space, characterized in that, The method includes: S1. Obtain standardized data on underground space; S2. Perform dynamic target detection on the standardized data to obtain the initial target information of the underground space; S3. Based on the multi-view geometric constraints of the underground space, evaluate the motion consistency of the initial target information under different viewpoints to establish a preliminary cross-view trajectory association of the underground space; S4. Based on the preliminary cross-view trajectory association, the initial target information is fused into a continuous trajectory under a unified coordinate system to generate the fused trajectory data of the underground space; S5. Project the fused trajectory data into a three-dimensional space, and perform a natural object fall source tracing analysis on the fused trajectory data to obtain the source location information of the underground space; S6. Integrate the source location information with the preset safety knowledge graph to generate visual data on the safety situation of the underground space.

2. The method for visualizing edge data of underground space security situation as described in claim 1, characterized in that, The acquisition of standardized data for underground space includes: Acquire multi-source monitoring data of underground space; The multi-source monitoring data is processed to unify the format, resulting in unified format data of the underground space. Redundant and invalid data are removed from the standardized data to obtain valid data for the underground space. By completing the key data fields of the valid data, the standardized data of the underground space is obtained.

3. The method for visualizing edge data of underground space safety situation as described in claim 1, characterized in that, The step of performing dynamic target detection on the standardized data to obtain initial target information for the underground space includes: Based on the safety monitoring requirements of the underground space, the dynamic target detection range of the underground space is defined; Target recognition is performed on the image data in the standardized data to capture target objects with motion characteristics in the underground space; The target object's morphological characteristics, time of appearance, and initial location in the monitoring field of view are integrated into basic attribute information; The basic attribute information is classified and organized to obtain the initial target information of the underground space.

4. The method for visualizing edge data of underground space security situation as described in claim 1, characterized in that, The evaluation of the motion consistency of the initial target information under different viewpoints based on the multi-view geometric constraints of the underground space, in order to establish a preliminary cross-view trajectory association of the underground space, includes: Based on the actual structural layout of the underground space, the relative positional relationships of the multi-view monitoring devices in the underground space are determined in order to establish the multi-view geometric constraints of the underground space. Identify the movement direction and rhythm trend of the same target object in the initial target information under different monitoring views to obtain the movement characteristics of the initial target information; Based on the multi-view geometric constraints, the degree of fit of the motion features under different monitoring views is evaluated to obtain the motion consistency of the initial target information; The target motion trajectory segments in the initial target information that meet the preset conditions for motion consistency are associated and matched to establish a preliminary cross-view trajectory association for the underground space.

5. The method for visualizing edge data of underground space safety situation as described in claim 4, characterized in that, The formula for calculating the degree of fit is as follows: ； In the formula, Indicates the degree of fit. This indicates the number of samples within a preset time window. The viewpoint in the multi-view geometric constraint condition represents the perspective. To view The fundamental matrix, Indicates time From the perspective of time The target location coordinates in the middle, Indicates time From the perspective of time The target location coordinates in the middle, This represents the preset scaling parameter used to normalize the polar error. An exponential function representing the natural constant.

6. The method for visualizing edge data of underground space security situation as described in claim 1, characterized in that, The step of fusing the initial target information into a continuous trajectory in a unified coordinate system based on the preliminary cross-view trajectory association to generate the fused trajectory data of the underground space includes: A unified spatial coordinate system for the underground space is established by using a pre-set fixed reference marker within the underground space as a reference. The trajectory segments in the preliminary cross-view trajectory association are mapped to the unified coordinate system to calibrate the spatial position of the trajectory segments, thereby obtaining the calibrated trajectory segments of the underground space; Determine the rationality of the connection between the trajectory segments in order to retain the trajectory connection relationships with high consistency among the trajectory segments; Based on the trajectory connection relationship, reasonable transition information at the connection point of the calibration trajectory segment is supplemented to obtain the fused trajectory data of the underground space.

7. The method for visualizing edge data of underground space safety situation as described in claim 1, characterized in that, The process of projecting the fused trajectory data into three-dimensional space and performing natural object fall source analysis on the fused trajectory data to obtain the source location information of the underground space includes: Based on the actual three-dimensional structural data of the underground space, determine the correspondence rules between the fused trajectory data and the physical location in the three-dimensional space; Based on the correspondence, the fused trajectory data is projected onto the three-dimensional space to obtain the three-dimensional trajectory data of the underground space; By filtering out the trajectory data corresponding to the natural object from the three-dimensional trajectory data, the motion trajectory fragments of the natural object are obtained; Based on the trajectory supplementary information recorded by the multi-view monitoring equipment in the underground space, the trajectory deviation of the natural object movement trajectory segment is corrected; By tracing the starting point of the corrected trajectory of the natural object and combining it with the position markers in the three-dimensional space, the starting spatial coordinates of the natural object's fall are determined. Based on the specific regional division of the underground space, the spatial positioning verification of the starting spatial coordinates is performed to obtain the source location information of the underground space.

8. The method for visualizing edge data of underground space security situation as described in claim 1, characterized in that, The process of integrating the source location information with a preset safety knowledge graph to generate visualized safety situation data for the underground space includes: Based on the spatial identifier in the source location information, match the corresponding spatial region and associated security rules in the preset security knowledge graph; Based on the safety rules, the types of natural objects corresponding to the source location information are associated with the types of hazards in the safety knowledge graph to determine the potential safety risk attributes of the underground space; Verify the logical consistency between the source location information and the potential security risk attributes to correct the source location information; Multimodal evolution is performed on the core security information in the corrected source location information to obtain the security situation visualization data of the underground space.

9. The method for visualizing edge data of underground space security situation as described in claim 8, characterized in that, The process of performing multimodal evolution on the core security information in the corrected source location information to obtain the security situation visualization data of the underground space includes: According to the preset multimodal presentation specifications, the text-based association information of the core security elements in the corrected source location information is converted into standardized explanatory text to obtain the standardized text data of the corrected source location information. Using preset graphical identification rules, the spatial association data in the core safety elements is graphically transformed to obtain the spatial graphical data of the underground space; Add time-series association markers to the standardized text data and the spatial graphical data to bind the state change trend of core security information with the time dimension, thereby obtaining the multimodal integrated data of the underground space; Based on the requirements of the underground space security situation display scenario, duplicate information elements in the multimodal integrated data are removed, and the information complementarity and logical coherence between different modal data are strengthened to obtain the underground space security situation visualization data.

10. A system for visualizing edge data of underground space safety situation, used to implement the method for visualizing edge data of underground space safety situation as described in claim 1, the system comprising: The data acquisition module is used to acquire standardized data of underground space; The target detection module is used to perform dynamic target detection on the standardized data to obtain the initial target information of the underground space; The trajectory association module is used to evaluate the motion consistency of the initial target information under different viewpoints based on the multi-view geometric constraints of the underground space, so as to establish a preliminary cross-view trajectory association of the underground space. The trajectory fusion module is used to fuse the initial target information into a continuous trajectory in a unified coordinate system based on the preliminary cross-view trajectory association, so as to generate the fused trajectory data of the underground space; The source tracing analysis module is used to project the fused trajectory data into a three-dimensional space and perform a natural object fall source tracing analysis on the fused trajectory data to obtain the source location information of the underground space; The visualization generation module is used to integrate the source location information with the preset safety knowledge graph to generate visualization data of the safety situation of the underground space.

Images