Multi-source data fusion train area perception and active protection system and method

Abstract

The invention relates to the technical field of railway transportation safety, in particular to a multi-source data fusion train area sensing and active protection system and method.The method comprises the steps that a data acquisition module obtains obstacle sensing data in front of a train and receives state and position data reported by a personnel terminal; the space-time reference and electronic fence management module provides unified positioning and map services and generates an electronic fence according to a plan; the multi-source situation fusion module carries out fusion processing on the information to generate a global dynamic situation; the real-time risk quantification module calculates a collision time TTC based on the situation, and outputs an evaluation result according to a dynamic risk model; the intelligent early warning decision-making module generates a graded early warning instruction containing specific action guidance; and finally, the early warning cooperative execution module synchronously distributes the instruction to the train and personnel terminal. According to the invention, the conversion from passive monitoring to active early warning is realized, and the safety protection capability of driving and operation in a complex environment is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of railway transportation safety technology, and in particular to a train regional perception and active protection system and method based on multi-source data fusion. BACKGROUND

[0002] Traditional railway safety protection mainly relies on a system with human defense as the core, that is, through the setting of on-site protection personnel, station liaison personnel and other posts, the use of whistles, flags, intercoms and other tools, and the combination of operation plans and train operation timetables for manual lookout, communication liaison and safety control; in recent years, with the progress of technology, some auxiliary technical means have also appeared, such as equipping operation personnel with alarm devices with satellite positioning function, or installing video monitoring equipment at fixed points, which to some extent provide personnel position information or visual monitoring capability in local areas.

[0003] Whether it is a traditional human defense system or the above-mentioned scattered technical auxiliary means, its essence still belongs to a discrete and passive monitoring mode; each sensing element is independent of each other, the information is fragmented, and there is a lack of effective fusion and intelligent analysis, resulting in a serious sensing blind area in complex operating environments, and a significant lag in the discovery and early warning of sudden conditions. SUMMARY

[0004] In order to make up for the above shortcomings, the present application provides a train regional perception and active protection system and method based on multi-source data fusion, aiming to improve the problem of a serious sensing blind area in complex operating environments, and a significant lag in the discovery and early warning of sudden conditions.

[0005] In a first aspect, the present application provides the following technical solution: a train regional perception and active protection system based on multi-source data fusion, comprising the following modules:

[0006] A data acquisition module for acquiring obstacle perception data of a region in front of a train running and receiving personnel state and position data reported by a field operation terminal;

[0007] A space-time reference and electronic fence management module for providing positioning data of a train and personnel, a railway line electronic map, and generating and managing an electronic fence on the electronic map according to an operation plan;

[0008] A multi-source situation fusion module for fusion processing of the obstacle perception data, personnel state and position data, electronic map and electronic fence to generate global dynamic situation information;

[0009] A real-time risk quantification module for calculating a time to collision (TTC) between a train and an operation personnel based on the global dynamic situation information, and outputting a risk assessment result according to the TTC and a preset dynamic risk threshold model;

[0010] an intelligent early warning decision module, configured to generate a hierarchical early warning decision instruction containing specific action guidance according to the risk assessment result;

[0011] an early warning collaborative execution module, configured to execute the early warning decision instruction and synchronously distribute early warning information to a train operation terminal and a work terminal.

[0012] By adopting the above technical solutions, a complete technical chain from multi-source fusion perception, global situation generation to dynamic risk assessment and collaborative early warning is realized, discrete and passive monitoring information is integrated into a unified and real-time panoramic situation view, and prospective risk quantification and hierarchical decision are made based on time to collision TTC, so that the perception blind area in a complex environment is systematically eliminated, and the starting point of safety early warning is greatly advanced from post-response to pre-prevention, thereby solving the problem of traditional early warning lag.

[0013] Preferably, the obstacle perception data acquisition includes:

[0014] The front area of the train operation is scanned by a radar to obtain original detection data containing distance, speed and direction information;

[0015] Image acquisition is performed on the area covered by the radar scanning to obtain visible light or infrared image data;

[0016] The original detection data and image data are fused and processed to identify and extract obstacle information in the track and surrounding area, and to generate obstacle perception data.

[0017] Preferably, the personnel state and position data reception includes:

[0018] The identity information, vital sign state data and position data based on the Beidou satellite navigation system of the work personnel are collected in real time through an intelligent safety terminal carried by the work personnel;

[0019] The identity information, state data and position data are uploaded to the system through a wireless communication network;

[0020] The uploaded data are received and analyzed, and are used as personnel state and position data.

[0021] Preferably, the generation and management of the electronic fence include:

[0022] The accessed work plan data are analyzed, and the position information and time window of the work section are extracted;

[0023] Based on the position information, a corresponding graphical area is delineated on a railway line electronic map to generate an electronic fence;

[0024] The time window is associated with the electronic fence, and the electronic fence is activated, continuously monitored and state updated.

[0025] Preferably, the fusion processing includes:

[0026] The obstacle perception data, the personnel state and position data are spatio-temporally aligned and unified to a spatio-temporal reference defined by the electronic map;

[0027] The electronic fence is used as a spatial constraint condition to associate the target in the obstacle perception data with the individual in the personnel state and position data;

[0028] Based on the association result, the train, the personnel, the electronic fence and the identified risk area are uniformly coded and integrated to generate global dynamic situation information.

[0029] Preferably, the calculation of the collision time TTC between the train and the operating personnel includes:

[0030] The real-time position and speed information of the train and the real-time position information of the operating personnel contained in the global dynamic situation information are acquired in real time;

[0031] Based on the real-time position information of the train and the operating personnel, the relative distance between them is calculated;

[0032] Based on the speed information of the train and the movement information of the operating personnel, the approaching speed of the train relative to the operating personnel is calculated;

[0033] According to the relative distance and the approaching speed, the collision time TTC is calculated.

[0034] Preferably, the dynamic risk threshold model includes:

[0035] According to the train type, the running speed and the line condition, the TTC threshold range corresponding to different risk levels is dynamically determined;

[0036] Based on the TTC threshold range, a mapping relationship from the TTC value to the risk level is established;

[0037] In real-time risk assessment, according to the calculated TTC value and the mapping relationship, the corresponding risk level is output.

[0038] Preferably, the generation of the graded early warning decision instruction includes:

[0039] According to the risk level in the risk assessment result, the corresponding instruction template in the preset early warning strategy library is matched;

[0040] Combined with the personnel position and the line environment in the global dynamic situation information, an action guide containing specific avoidance direction, recommended operation intensity and estimated remaining time is generated;

[0041] The action guide is encapsulated with the corresponding early warning level to generate a hierarchical early warning decision instruction.

[0042] Preferably, the early warning information distribution includes:

[0043] The early warning decision instruction is converted into a communication protocol and data format suitable for the train operation terminal and the operation terminal respectively to generate an early warning data packet to be distributed;

[0044] The early warning data packet to be distributed is synchronously sent to the corresponding train operation terminal and operation terminal through a railway special wireless communication network;

[0045] The terminal triggers the output of early warning prompts through at least two channels of visual display interface, auditory alarm device and tactile vibration device according to the early warning data packet.

[0046] In a second aspect, the present application provides the following technical solutions: a train regional perception and active protection method based on multi-source data fusion, the method comprising:

[0047] Obtain obstacle perception data of the area in front of the train operation, and receive personnel state and position data reported by the field operation terminal;

[0048] Provide positioning data of the train and personnel, a railway line electronic map, and generate and manage an electronic fence on the electronic map according to the operation plan;

[0049] Fuse the obstacle perception data, personnel state and position data, electronic map and electronic fence to generate global dynamic situation information;

[0050] Based on the global dynamic situation information, calculate the collision time TTC between the train and the operation personnel, and output the risk assessment result according to the TTC and a preset dynamic risk threshold model;

[0051] According to the risk assessment result, generate a hierarchical early warning decision instruction containing specific action guide;

[0052] Execute the early warning decision instruction to synchronously distribute early warning information to the train operation terminal and the operation terminal.

[0053] The present application has the following beneficial effects:

[0054] 1、In the present application, by integrating vehicle-mounted radar-vision fusion perception, personnel intelligent terminal reporting, Beidou high-precision positioning and electronic map, a perception network of real-time synchronization of multi-source information is constructed, three-dimensional and all-weather monitoring of the train operation environment and the state of the operation personnel on the railway line is realized, and the problems of coverage blind area, reaction lag and insufficient reliability existing in traditional manual lookout and single technical means are solved from the data source.

[0055] 2、In the application, the multi-source heterogeneous data is deeply fused and spatio-temporal registered by the central processing module to generate unified and coherent global dynamic situation information, so that the dispatch personnel can master the dynamic relationship between train position, personnel distribution, electronic fence and risk area on the integrated electronic map in real time, realize the transformation from dispersed information to centralized situation, and provide unprecedented panoramic view support for accurate safety decision.

[0056] 3、In the application, the risk assessment is carried out by using the time to collision TTC and dynamic risk threshold model based on real-time calculation, which replaces the traditional fixed distance alarm mode, so that the system can intelligently adjust the warning level and timing according to the real-time speed of the train, braking performance and relative motion relationship, realizes the technical leap from static threshold to dynamic self-adaption in risk judgment, and improves the scientificity and pertinence of the warning. BRIEF DESCRIPTION OF DRAWINGS

[0057] Figure 1 The architecture diagram of the train regional perception and active protection system of multi-source data fusion provided by the application;

[0058] Figure 2 The flowchart of the train regional perception and active protection method of multi-source data fusion provided by the application. DETAILED DESCRIPTION

[0059] The technical solutions in the embodiments of the application will be described clearly and completely below with reference to the drawings. Obviously, the described embodiments are only part of the embodiments of the application, not all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor are within the protection scope of the application.

[0060] Embodiment one:

[0061] In the first embodiment of the application, the application provides a train regional perception and active protection system of multi-source data fusion, as shown in Figure 1 The system includes the following modules:

[0062] The data acquisition module is used to acquire obstacle perception data of the region in front of the train running, and receive personnel state and position data reported by the field operation terminal;

[0063] Further, the acquisition of obstacle perception data includes:

[0064] The region in front of the train running is scanned by radar to obtain original detection data containing distance, speed and direction information;

[0065] The image acquisition is carried out on the region covered by the radar scanning to obtain visible light or infrared image data;

[0066] The original detection data and image data are fused to identify and extract obstacle information in the track and surrounding area, and generate obstacle perception data.

[0067] Further, the received personnel state and position data includes:

[0068] Through the intelligent safety terminal carried by the operating personnel, real-time collection of identity information, vital sign state data and position data based on the Beidou satellite navigation system is realized;

[0069] The identity information, state data and position data are uploaded to the system through the wireless communication network;

[0070] The uploaded data is received and analyzed, and is used as personnel state and position data.

[0071] Specifically, the data acquisition module is configured to obtain key operation information facing both the train operation environment and the field operating personnel; the module runs in the cooperative system of the vehicle-mounted control unit and the back-end server, and through unified access and management of sensor input, operation terminal input and communication link input, the subsequent situation fusion, risk calculation and early warning decision-making have a continuous, reliable and analyzable data basis;

[0072] When obtaining the obstacle perception data of the area in front of the train operation, the radar acquisition subunit first starts to scan the space in front of the line; one possible way is to use a millimeter wave radar or a laser radar to perform periodic scanning, and generate original detection data containing target distance, radial velocity and azimuth angle during the scanning process; these original detection data are usually recorded in polar coordinate format, providing a unified input structure for subsequent spatial projection and recognition algorithms; within the synchronization period when the radar starts scanning, the image acquisition subunit performs image acquisition on the radar coverage area; optionally, the image acquisition can include visible light image, infrared image or a combination of the two, and generate a continuous image sequence through a fixed frame rate; in order to ensure the matchability of the image data and the radar detection data, the image acquisition subunit adopts a time synchronization mechanism consistent with the radar scanning period, for example, using a unified time signal as the clock reference, so that each frame of image can be matched to the corresponding radar measurement period;

[0073] After the collection is completed, the data analysis and buffer subunit inputs the original detection data and image data into the fusion processing flow; in this flow, the system first completes the preliminary matching of the two types of data according to the time stamp, and then establishes the mapping relationship between the radar coordinate system and the image coordinate system through the spatial calibration parameter; one possible mapping method is to project the radar polar coordinate data to the image plane by using the external parameter matrix, so as to locate the spatial region corresponding to the radar echo in the image; in the unified coordinate frame, the system identifies the track area and the potential obstacles around the track by using the joint strategy of image feature detection and radar point cloud clustering; in this identification process, each detection target can be evaluated by constructing a joint confidence function, and the confidence function may be expressed as:

[0074] ;

[0075] wherein, represents the target confidence generated by the radar feature, represents the target confidence generated by the image feature, and is a weight parameter for adjusting the contribution of the two types of features to the final identification result; after the fusion calculation is completed, the system outputs the obstacle perception data containing the position, speed, shape type and confidence of each obstacle, and stores it in the buffer area for calling by the situation fusion module;

[0076] In the process of receiving personnel state and position data, the terminal access subunit maintains connection with the intelligent safety terminal carried by the operating personnel through the wireless communication link; the intelligent terminal can monitor the personnel identity information, vital sign state and position information based on the Beidou satellite navigation system; after the terminal collects the above data, it is uploaded to the system through the special wireless communication network commonly used in the railway site according to the preset data upload period; after receiving the uploaded data, the terminal access subunit checks the integrity of the data packet, the time stamp, the identity mark and the positioning accuracy information; after the data verification is passed, the data analysis and buffer subunit analyzes it into structured personnel state and position records; these records usually include the unique identity of the personnel, the vital sign parameters such as heart rate or motion state, the Beidou positioning coordinates, the collection time stamp and the terminal running state information, which are stored in a unified format to ensure that the obstacle perception data can be synchronized in the subsequent processing process;

[0077] In the process of the overall operation of the data acquisition module, various data from the radar, images and the terminal of the operating personnel are written into the cache in a continuous stream manner, and are attached with standardized time labels and positioning labels; these data are then pushed to the situation fusion module at a configurable refresh frequency, so that the system can construct a dynamic situation based on the latest environmental conditions and personnel states; the design of the module ensures that the information of the vehicle end and the personnel end is continuously acquired, seamlessly connected and computationally expressed in the system, thereby providing a data basis for subsequent TTC calculation, risk judgment and warning output.

[0078] The space-time reference and electronic fence management module is configured to provide positioning data of the train and the personnel, an electronic map of the railway line, and generate and manage an electronic fence on the electronic map according to the operation plan;

[0079] Further, the generation and management of the electronic fence include:

[0080] The accessed operation plan data is parsed to extract location information and a time window of the operation section;

[0081] Based on the location information, a corresponding graphical area is delineated on the electronic map of the railway line to generate the electronic fence;

[0082] The time window is associated with the electronic fence, and the electronic fence is activated, continuously monitored and managed in a state updating manner.

[0083] Specifically, the space-time reference and electronic fence management module manages the spatial positions of the train and the operating personnel by accessing the positioning data, the electronic map of the railway line and the operation plan information, and generates a dynamically updateable electronic fence on the electronic map according to the operation activities; the module is usually operated in a central server or a dispatching control platform, and performs data interaction with the positioning service, the operation plan system and the situation fusion module through a standardized interface, so as to ensure that the electronic fence is consistent and computationally expressible in the time dimension and the space dimension;

[0084] In processing the operation plan data, the module first parses the received plan file or plan interface data; the operation plan generally includes the line range, the mileage coordinate or the latitude and longitude coordinate of the operation section, and the plan start time and the plan end time; in order to ensure that the data from different sources can be used in cooperation with the electronic map, the module formats the section location parameters and takes the projection coordinate system used by the electronic map as a unified space reference; for example, when the input is latitude and longitude, the geographic coordinates are optionally mapped to plane coordinates through a projection transformation function , and the mapping relationship can be expressed as:

[0085] ;

[0086] wherein, latitude, longitude, projection function; the mapping ensures that the work section can be accurately marked in the spatial framework of the electronic map, providing a unified expression for subsequent fence boundary generation; in the time dimension, the module standardizes the planning time window and synchronizes the time stamp with a unified time source to avoid inconsistencies in fence activation timing caused by different time systems;

[0087] In the process of generating an electronic fence, the module draws a graphical area on the map according to the spatial range of the work section; the boundary of the fence can be automatically expanded by a certain width according to the work area center line to cover the personnel activity space and the safety buffer distance; the buffer width can be set according to rules, for example, determined by a basic buffer value and optional environmental parameters; the module constructs a polygon boundary in order from the processed coordinate points and performs topological checking on the geometric shape to ensure that it forms a legal closed area; after generating the fence layer, the module stores it as a retrievable map entity and attaches attribute fields such as section identification, work type, and plan number for subsequent management and dispatching system calls;

[0088] In the association process of the fence and the time window, the module establishes a life cycle management record for each electronic fence, including fence ID, fence boundary, plan start time , plan end time , optional early activation buffer time and late retention time, and the current running state of the fence; the actual activation of the fence is judged according to the relationship between the current system time and the planning time window; when is satisfied, the fence enters the active state; in the above symbols, represents the current system time, represents the start time of the work, represents the end time of the work; in the active state, the module will continuously monitor the fence-related events, including whether the train or personnel enters the fence boundary and whether the fence state needs to be adjusted according to the plan change;

[0089] In the process of continuous monitoring and status update, the module maintains the active state of the fence in an event-driven or timed task manner; when the positioning module provides the train position or personnel position, the module performs a position judgment algorithm on the position point to determine whether the target falls into the fence area; the judgment is usually based on a geometric algorithm for points inside a polygon, and is calculated by the intersection number or ray method of a planar polygon; once an entry event or a departure event is detected, the module generates an event record and reports it to the situation fusion module for subsequent risk assessment and early warning decision; in the case of fence coverage conflict, plan change or positioning anomaly, the module will adjust the fence state according to the preset strategy and synchronize it to other functional modules through the event interface;

[0090] In practical applications, the space-time reference and electronic fence management module provides the system with continuously updated spatial operation boundaries, so that the real-time positions of trains and personnel can be analyzed in a unified map framework; after the fence is activated, the module can provide visual display of the fence to the train end device and the operation terminal, and support low-bandwidth loading of mobile terminals through a lightweight data interface; at the end of the fence life cycle, the module archives relevant boundary data, event records and life cycle information for subsequent audit or security analysis; through the above mechanism, the module ensures the whole process operation of the electronic fence from generation, activation to monitoring, and provides a reliable spatial basis and time basis for subsequent multi-source situation fusion and risk quantification.

[0091] The multi-source situation fusion module is used for fusion processing of obstacle perception data, personnel state and position data, electronic map and electronic fence, and generates global dynamic situation information;

[0092] Further, the fusion processing includes:

[0093] The obstacle perception data and the personnel state and position data are spatio-temporally aligned and unified to the space-time reference defined by the electronic map;

[0094] The electronic fence is used as a spatial constraint condition to associate the targets in the obstacle perception data with the individuals in the personnel state and position data;

[0095] Based on the association result, the train, personnel, electronic fence and identified risk area are uniformly coded and integrated to generate global dynamic situation information.

[0096] Specifically, the multi-source situation fusion module processes the multi-source heterogeneous data from the obstacle perception module, the personnel terminal, the positioning module, and the electronic fence management module to construct complete, continuous, and quantifiable global dynamic situation information; the module usually runs on the server side or edge computing nodes, and through the spatio-temporal alignment, spatial correlation, and unified coding of data from different sources, the train, personnel, obstacles, and fence areas are integrated into a unified spatio-temporal framework to provide the computable description required by the subsequent risk assessment module; the module undertakes the core function of data integration in the system structure, and its running mechanism directly affects the accuracy and real-time performance of the early warning decision;

[0097] In the spatio-temporal alignment, the multi-source situation fusion module first preprocesses the obstacle perception data and personnel state and position data; since the two types of data come from vehicle-mounted radars, image acquisition devices, and personnel terminals, their time stamp formats, sampling periods, and coordinate systems may differ, so the module establishes a standardized time reference through a unified time source, taking the current system time or time signal as the global time reference; in the time axis alignment, the module performs linear interpolation or nearest neighbor matching according to the data time stamp to align data with different sampling periods to a common time scale; in the spatial alignment process, the module takes the projection coordinate system used by the electronic map as the unified spatial reference, and maps the spatial position in the obstacle data from the radar coordinate system or image coordinate system to the map coordinate system through the external parameter matrix; the mapping can be represented as:

[0098] ;

[0099] wherein, represents the point coordinate in the original coordinate system, represents the point coordinate in the electronic map coordinate system, is a two-dimensional homogeneous transformation matrix containing rotation, translation, and scale parameters; through the transformation, the obstacle position, personnel position, and train position can form a unified spatial semantic expression on the electronic map;

[0100] After completing the spatio-temporal alignment, the module correlates the obstacle perception data and personnel position data with the electronic fence as a spatial constraint; the electronic fence appears as a polygon area on the electronic map, so the module uses the point-in-polygon method to determine whether the target is inside the fence; for example, the ray casting method is used to determine whether the target point whether the polygon inclusion condition is met; in the judgment process, the module not only identifies whether the personnel are inside the fence, but also identifies whether the obstacle target is located near the fence boundary or crosses the fence area, thereby forming the spatial relationship between the target and the fence; the spatial constraint can clearly define the risk range of the work, so that the subsequent risk model can quantify the relative position relationship between the train and the personnel based on the fence boundary; for the spatial relationship between the train and the obstacle, and the train and the personnel, the module further calculates the relative distance For risk area identification, the distance can be calculated according to the Euclidean distance formula:

[0101] ;

[0102] wherein, is the train position coordinate, is the personnel or obstacle position coordinate; the distance provides a basic quantitative index for subsequent identification of risk areas and estimation of collision risks;

[0103] After the target association is completed, the module integrates the train, personnel, electronic fence and risk area based on the association results; the unified coding is to enable the downstream module to quickly analyze the input data, so the module establishes a standardized field set for each type of object, such as object type identification, spatial coordinates, motion speed, state label, risk level placeholder field and relationship label with the fence; in terms of risk area identification, the module marks potential risk areas according to the distance threshold, target classification results and fence geometric relationship, such as marking the obstacle near the train track or the personnel near the fence boundary as a risk object; in the integration process, the module combines all object coding structures into a unified dynamic situation message, which contains multiple object nodes and their spatial and semantic relationships, so that the situation information is expressed in the form of a graph structure and can be used for real-time updating;

[0104] In the final stage of generating global dynamic situation information, the module organizes and caches the aforementioned coding results, so that the system can push the situation information to the risk quantification module in a fixed period or event triggered manner; the global dynamic situation information is usually presented in the form of a structured data packet, which contains time label, train node, personnel node, obstacle node, fence layer and risk area list; in the application layer, the dynamic situation information can be used by the early warning decision module to formulate instructions, and can also be used for interface display to realize visual situation presentation; under this working mechanism, the multi-source situation fusion module integrates multiple types of heterogeneous data into a unified and computable environment representation, so that the system can perform subsequent risk analysis and early warning generation based on real-time situation.

[0105] The real-time risk quantification module is configured to calculate a time-to-collision (TTC) between the train and the worker based on the global dynamic situation information, and output a risk assessment result according to the TTC and a preset dynamic risk threshold model.

[0106] Further, the calculation of the TTC between the train and the worker includes:

[0107] The real-time global dynamic situation information includes real-time position and speed information of the train, and real-time position information of the worker.

[0108] Based on the real-time position information of the train and the worker, a relative distance between the train and the worker is calculated.

[0109] Based on the speed information of the train and the movement information of the worker, an approaching speed of the train relative to the worker is calculated.

[0110] According to the relative distance and the approaching speed, the TTC is calculated.

[0111] Further, the dynamic risk threshold model includes:

[0112] According to the train type, the running speed, and the line condition, TTC threshold ranges corresponding to different risk levels are dynamically determined.

[0113] Based on the TTC threshold ranges, a mapping relationship from the TTC value to the risk level is established.

[0114] In real-time risk assessment, according to the calculated TTC value and the mapping relationship, the corresponding risk level is output.

[0115] Specifically, the real-time risk quantification module continuously calculates the TTC between the train and the worker based on the global dynamic situation information, and outputs a risk assessment result according to a preset dynamic risk threshold model. The module is deployed in a central processing unit or an edge computing node, and maintains a low-delay data channel with the situation fusion module, the electronic fence management module, and the early warning decision module, so as to ensure that the risk calculation can be completed in a short period after data update and to promote subsequent early warning actions.

[0116] The module first acquires data items from the situation information in real time, including a real-time position vector and a speed vector of the train, and a real-time position vector and a speed vector of the worker. Time is denoted as t; to ensure the accuracy of the calculation, the input data needs to be processed by filtering and interpolation before use, such as using Kalman filtering or weighted smoothing to denoise and short-term predict the speed and position, so as to obtain more stable estimated values; based on the position vector, the module calculates the relative distance between the train and the personnel and expressed in Euclidean norm, that is,

[0117] ;

[0118] wherein represents the position coordinates of the train in the map coordinate system, represents the position coordinates of the personnel;

[0119] After obtaining the relative position, the module calculates the closing speed of the train relative to the personnel ; One possible calculation method is to project the relative speed vector in the direction of the relative position to obtain the scalar closing rate, which is specifically expressed as:

[0120] ;

[0121] wherein, when , it means that they are approaching, and when , it means that it does not constitute a convergence risk; to avoid short-term misjudgment caused by noise, the module can adopt a short-term maximum or sliding average strategy for robust processing of ;

[0122] After obtaining the relative distance and the closing speed , the module calculates the time to collision TTC, and the basic mathematical form is:

[0123] ;

[0124] wherein is the minimum effective threshold of the closing speed, which is used to filter measurement noise and relative static situation, and the symbol meaning is: is the relative distance, is the closing speed, is the estimated remaining collision time; when is less than or equal to , the module regards TTC as infinite to represent no collision trend in the short term;

[0125] In order to compare TTC with the train operation safety capability, the module can further introduce the estimation of train braking time and stopping distance to form a more comprehensive risk judgment; one possible braking time estimation is to calculate the braking time and braking distance based on the current train speed scalar and the braking deceleration parameter , which are With ; where is the typical braking deceleration achievable by the train under current conditions, the value can be parameterized by train type and line condition; the module can compare TTC with and the corresponding time margin, when or , it is marked as extremely high risk or emergency level, where and are safety margin parameters set by human to consider response delay and uncertainty;

[0126] The risk threshold model adopts a dynamic threshold strategy to reflect the differences in train type, running speed and line condition; the threshold model calculates a set of risk level thresholds according to input parameters such as train category , current speed and line factors such as curve radius, slope, visibility influence coefficient, which defines the mapping relationship from TTC to risk level, which can be formally represented as:

[0127]

[0128] where , and are configurable threshold functions, the module can be initialized by empirical data or train braking performance parameters, and allow adjustment by regulatory agencies or experts during operation and maintenance phase;

[0129] In actual operation, the risk threshold function can be parameterized, for example, let the threshold be inversely proportional to train speed to reflect the demand for early warning of high-speed trains, which can be formally represented as:

[0130] ;

[0131] where the subscripts and are configurable coefficients based on train type and line condition; this expression ensures that the corresponding TTC threshold shrinks when the train speed rises, thus triggering an earlier warning; when deployed, the coefficients are calibrated by authoritative test data or train manufacturer-provided braking performance parameters to ensure that the threshold is engineering-achievable;

[0132] To enhance robustness, the module will perform confidence weighting and fusion judgment before outputting the risk assessment result; the confidence can come from the positioning accuracy, speed estimation variance and sensor fusion confidence term, the module will package TTC and confidence into a risk assessment message and send it to the early warning decision module, the risk message contains field examples such as: timestamp, train ID, personnel ID, TTC value, approach speed, risk level, confidence score and trigger factor description; in the case of confidence below threshold or data missing, the module will issue a data quality alarm and trigger a degradation strategy, such as strengthening remote manual monitoring or using conservative thresholds for early warning;

[0133] Finally, the module will issue the risk assessment result in the form of event or periodic package to the early warning decision module and monitoring center according to the configured release strategy, for driving hierarchical early warning, voice prompt and necessary linkage control, so as to realize the closed loop from real-time situation to executable early warning.

[0134] The intelligent early warning decision module is configured to generate a hierarchical early warning decision instruction containing specific action guidance according to the risk assessment result.

[0135] Further, generating the hierarchical early warning decision instruction comprises:

[0136] According to the risk level in the risk assessment result, matching the corresponding instruction template in the preset early warning strategy library;

[0137] Combining the personnel position and line environment in the global dynamic situation information, generating an action guidance containing specific avoidance direction, recommended operation intensity and estimated remaining time;

[0138] The action guidance and the corresponding early warning level are packaged to generate a hierarchical early warning decision instruction.

[0139] Specifically, the intelligent early warning decision module receives the risk assessment result from the real-time risk quantification module and automatically generates a hierarchical early warning decision instruction containing specific action guidance based on the preset strategy library; the module runs on the central processing unit and keeps low-delay communication with the situation fusion module, vehicle terminal and personnel terminal communication module, so as to timely issue executable instructions when risks occur;

[0140] The intelligent early warning decision module first performs semantic analysis and priority sorting on the incoming risk assessment record, the analysis content includes risk level, TTC value, approach speed, trigger object identification, associated fence ID, position information and confidence score; after analysis, the matching instruction template is retrieved in the early warning strategy library according to the risk level, the strategy library is a template rule library, each template contains instruction header, risk response category, recommended action set, communication adaptation format and optional manual intervention prompt; the template matching adopts risk level direct indexing or nearest neighbor matching based on score, to ensure that when the level is the same but the scene is different, the template that is more suitable for the scene can be selected;

[0141] In generating specific action guidance in combination with global dynamic situation information, the module reviews personnel position, train trajectory, fence geometry, line environmental factors, and current train operation parameters; based on these elements, the decision logic first calculates an avoidance direction suggestion, which can be output by the path planning sub-logic, with the goal of guiding personnel to the safest point farthest from the train movement direction and outside the electronic fence, with path planning scoring and outputting direction vectors based on passable areas of the electronic map and distance to the nearest safe exit ; for the suggestion on the train driver side, the module evaluates braking or deceleration strategies and gives a recommended operation strength, such as suggesting deceleration to a target speed or executing a braking strength , the recommended braking acceleration can be calculated to meet the target speed in the remaining distance :

[0142] ;

[0143] wherein represents the current train speed scalar, represents the suggested passing speed, represents the shortest distance along the track between the train and the risk point; the symbol meanings in the formula are issued together in the decision message, and limits the upper and lower bounds to adapt to the train braking capability and passenger comfort constraints;

[0144] In estimating the estimated remaining time, the module generates a more operational remaining time prompt based on the calculated TTC and in combination with train braking time and personnel response time; for example, the module can output the train arrival remaining time , and in combination with the average response time of personnel executing avoidance to give an estimated remaining avoidance time , wherein can be adaptively estimated based on terminal type, personnel state identification, and historical response statistics; the remaining time is displayed in the form of a countdown on the personnel terminal and broadcast in the form of a voice prompt at the train end;

[0145] For the language and format generation of action guidance, the module assembles template text and dynamic parameters into a directly deliverable instruction package, which includes instruction ID, target object list, action type, avoidance direction vector or text description, suggested operation strength parameter, estimated remaining time, confidence score, and traceability field; the instruction package also includes communication adaptation fields to generate train end format, personnel terminal simplified format, and monitoring center detailed format for the same instruction according to different protocol and bandwidth requirements; in low bandwidth situations, the personnel terminal receives a simplified instruction package containing only key text, countdown, and avoidance direction code;

[0146] In generating the decision instruction, the module also considers the confidence and multi-source consistency. When the risk assessment confidence is low or there is a sensor conflict, the decision logic triggers a degradation or redundancy strategy, such as selecting a more conservative With the increase Estimation, or simultaneously issue an artificial confirmation instruction to request the dispatcher or site manager to confirm; when the confidence is high and the trigger condition reaches the emergency level, the module will enable the strong constraint template, in addition to issuing instructions, it can also trigger the linkage interface to suggest or request the train control system to execute automatic deceleration or emergency braking. The associated linkage must comply with the safety certification requirements of the train control system and record the corresponding approval and execution logs;

[0147] To ensure the traceability and auditability of the instruction, the module numbers and records each generated decision instruction, the record content includes the original data reference that triggers the risk assessment, template version number, dynamic parameters, target and timestamp of the issued, and the final state feedback of the instruction; At the same time, the module supports manual intervention channels, allowing dispatchers or safety management personnel to revise or revoke automatically generated instructions on the monitoring interface. All manual operations are included in the audit chain for post-analysis;

[0148] In the application layer, the generated hierarchical early warning decision instruction is issued to the vehicle display and voice system, personnel terminal, and remote monitoring center; the personnel terminal outputs prompts in multiple channels and displays the avoidance direction and countdown, the vehicle display reminds the driver in the form of a map and voice and gives the recommended speed or braking parameters, and the monitoring center receives detailed instruction packages for manual scheduling and event recording.

[0149] The early warning coordination execution module is used to execute the early warning decision instruction and synchronously distribute the early warning information to the train operation terminal and the operation terminal.

[0150] Further, the distribution of early warning information includes:

[0151] Converting the early warning decision instruction into a communication protocol and data format suitable for the train operation terminal and the operation terminal respectively to generate a to-be-distributed early warning data package;

[0152] Synchronously sending the to-be-distributed early warning data package to the corresponding train operation terminal and operation terminal through a railway special wireless communication network;

[0153] The terminal triggers at least two channels of visual display interface, auditory alarm device and tactile vibration device to output early warning prompts according to the early warning data package.

[0154] Specifically, the early warning coordination execution module, as the execution layer for early warning issuance and terminal execution, receives hierarchical early warning decision instructions from the intelligent early warning decision module and is responsible for reliably and synchronously transmitting them to the train operation terminal and work terminal. It is also responsible for collecting and forwarding status feedback to support instruction closure and post-event auditing. This module is deployed on the central dispatch server or edge gateway node and interfaces with the railway dedicated wireless communication network and terminal equipment through three types of logical units: protocol adaptation layer, transmission scheduling layer, and terminal adaptation layer, thereby realizing format conversion, communication guarantee, and execution monitoring for different terminal types.

[0155] When converting early warning decision commands into communication protocols and data formats adapted to train operation terminals and work terminals, the module first performs semantic parsing on the received commands, extracting fields such as command ID, target object list, priority, timestamp, action guidance parameters, and confidence level. After parsing, the protocol adaptation layer selects an appropriate encapsulation format based on the capabilities of the target terminal and the communication channel. For example, for train operation terminals, it can encapsulate them into binary data packets conforming to the onboard control information format or command sequences conforming to the train control system interface specifications. For work terminals, it can generate lightweight JSON or binary compressed packets to adapt to the processing capabilities and bandwidth constraints of mobile terminals. During the encapsulation process, the module sets the priority bit in the message header according to the command priority and adds the expiration time TTL and retry policy parameters so that the transmission scheduling layer can schedule retransmission or degrade transmission when there is network congestion or link abnormality.

[0156] When generating early warning data packets to be distributed, the module encrypts and protects the integrity of the data packets to meet the confidentiality and anti-tampering requirements of railway safety communication. One possible approach is to use a symmetric key encryption algorithm to encrypt the message body and use a message authentication code for integrity verification. The message header contains an encryption algorithm identifier, a message version number, and a signature digest pointer. Meanwhile, to reduce bandwidth consumption, the module supports multi-level encoding strategies. For messages sent from the train end, high-precision graphics and parameterized commands can be used, along with map tile indexes for local rendering on the train end. For messages sent from the operation terminal, simplified content such as text instructions, avoidance direction codes, and countdown numbers is used. Before leaving the network, the data packets are packaged into a sending queue by the transmission scheduling layer, and the sending order is determined according to message priority and network conditions.

[0157] When synchronously sending early warning data packets to be distributed to corresponding terminals via the railway dedicated wireless communication network, the transmission scheduling layer selects the sending interface based on the network link type and executes a multi-path redundancy sending strategy to improve reliability. Synchronous sending can be achieved by simultaneously sending data packets to both the train terminal and the personnel terminal, requiring confirmation receipts from both ends within a logical time window to determine successful synchronization. If the confirmation receipt reception is delayed... Exceeding the preset threshold This will trigger a retransmission or a switch of the communication path; where denotes the time delay from sending to receiving the first valid receipt, is the maximum allowed receipt delay configurable by the system; to avoid blocking caused by short network congestion, the module supports the strategy of sending in batches in parallel with single emergency priority sending, and triggers targeted area broadcast or multicast when necessary to ensure coverage;

[0158] In the presence of abnormal situations such as weak coverage, limited bandwidth or packet loss in the network or terminal, the module executes a downgrade delivery strategy to ensure the reachability of critical prompts; the downgrade strategy includes text replacement, repeated short warning content, multiple frequency short packet retransmission and parallel sending in multiple modes, and an acknowledgment mechanism is adopted for important messages, i.e. the terminal is required to return an acknowledgment packet containing the execution status after receiving, if the terminal does not return a valid acknowledgment within the number of retries , the module will mark the event as a communication exception and alarm the monitoring center, the parameters can be configured to adapt to different network environments;

[0159] In the execution logic of triggering the terminal to output the early warning prompt, the terminal adaptation layer automatically drives at least two channels to prompt according to the action guidance parsed from the received data packet, one possible combination is to trigger the visual display interface and the auditory alarm device in parallel, or to trigger the visual and tactile vibration in parallel; the visual channel includes elements such as map highlighting, arrow indication, countdown number and text description, the auditory channel includes hierarchical voice broadcast or hierarchical bee signal, and the tactile channel is short-time strong vibration to ensure that people can perceive the warning in a noisy environment; after receiving the instruction, the terminal prioritizes the prompt channel and marks the instruction source and the remaining valid time on the display, so that the on-site personnel can quickly interpret and execute, and at the same time the terminal returns the execution feedback to the center through the uplink channel while executing the instruction;

[0160] During execution, the early warning cooperative execution module provides a closed-loop confirmation and exception handling mechanism, the center waits for the terminal's receipt after issuing the instruction and analyzes the status of the receipt to determine whether additional instructions or manual intervention are needed; when the receipt indicates that the personnel have not responded as instructed or the train end has not completed the expected deceleration, the module can automatically upgrade the warning level and issue stronger instructions again or send a request for manual intervention to the monitoring center at the same time; all issued and feedback events are recorded in detail, and the record fields include message ID, sending time, terminal ID, receipt time, receipt status and retransmission number, to facilitate subsequent tracing and auditing;

[0161] At the application level, the early warning cooperative execution module not only undertakes the task of immediate early warning distribution, but also provides configuration of delivery strategy, terminal capability abstraction and state monitoring function, operation and maintenance personnel can view the message sending queue, link quality statistics, terminal response rate and historical event playback through the management interface, and can forcibly issue manual voice or broadcast instructions through the interface in emergency situations.

[0162] Embodiment two:

[0163] When the line maintenance operation is carried out in the low-visibility environment such as mountain railway curve or night rain and fog, the traditional protection means is blocked by terrain and climate, there are serious visual blind area and perception delay, it is difficult to find the approaching train and give early warning. In order to solve the above problems, the train regional perception and active protection method of multi-source data fusion is adopted, and the structure is as shown in Figure 2 The specific implementation process is as follows:

[0164] Obtain the obstacle perception data of the area in front of the train running, and receive the personnel state and position data reported by the field operation terminal;

[0165] Provide the positioning data of the train and personnel, the electronic map of the railway line, and generate and manage the electronic fence on the electronic map according to the operation plan;

[0166] Fuse the obstacle perception data, personnel state and position data, electronic map and electronic fence, and generate global dynamic situation information;

[0167] Based on the global dynamic situation information, calculate the collision time TTC between the train and the operation personnel, and output the risk assessment result according to the TTC and the preset dynamic risk threshold model;

[0168] According to the risk assessment result, generate a hierarchical early warning decision instruction containing specific action guidance;

[0169] Execute the early warning decision instruction, and synchronously distribute the early warning information to the train operation terminal and the operation terminal.

[0170] Specifically, the fusion perception unit carried by the train obtains the front obstacle perception data, which can fuse millimeter wave radar and visual sensor data to overcome the limitations of single sensor; at the same time, the personnel state data containing Beidou high-precision position actively reported by the operation personnel intelligent terminal through wireless network is received;

[0171] Using the pre-set high-precision railway electronic map and Beidou positioning data, consistent coordinate reference is provided for all moving targets; at the same time, according to the operation plan that has been accessed, the electronic fence bound with the operation space-time range is automatically generated on the electronic map, which provides spatial and temporal constraints for subsequent risk judgment;

[0172] The obstacle data, the personnel data and the electronic fence are time-space aligned and associated, a working personnel and other risk targets located in the activated electronic fence are identified, and are integrated into a unified global dynamic situation information; the information completely describes real-time spatial relationships among the vehicle, the personnel, the ground and the plan, thereby constructing a global situation awareness capability beyond the naked eye vision at the system level;

[0173] is a calculation of a collision time TTC between the train and the working personnel, the system outputs a quantitative risk assessment level according to the calculated TTC value and a risk threshold model dynamically set based on a safety braking distance and other parameters;

[0174] According to the risk assessment level, a preset strategy library is matched to generate a hierarchical early warning decision instruction including a specific avoidance direction and an operation guide, so that the early warning information has explicit operability;

[0175] The early warning instruction is synchronously distributed to a train driving terminal and a related working personnel terminal, and the terminal is driven to perform warning through multiple redundant channels such as sound, light and touch; thereby realizing a complete closed loop from early risk perception, intelligent assessment to explicit warning execution, and effectively solving the early warning lag problem in a complex environment.

[0176] Finally, it should be noted that: the above only describes the preferred embodiments of the present application and is not intended to limit the present application, although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions recorded in the foregoing embodiments or make equivalent replacement for some technical features, any modification, equivalent replacement, improvement, etc. within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A train regional perception and active protection system with multi-source data fusion, characterized in that, The system comprises the following modules: a data acquisition module for acquiring obstacle perception data of an area in front of a train operation and receiving personnel state and position data reported by a field operation terminal; a space-time reference and electronic fence management module for providing positioning data of the train and personnel, an electronic map of a railway line, and generating and managing an electronic fence on the electronic map according to an operation plan; a multi-source situation fusion module for fusion processing of the obstacle perception data, the personnel state and position data, the electronic map and the electronic fence to generate global dynamic situation information; a real-time risk quantification module for calculating a time to collision (TTC) between the train and the operation personnel based on the global dynamic situation information, and outputting a risk assessment result according to the TTC and a preset dynamic risk threshold model; an intelligent early warning decision module for generating a hierarchical early warning decision instruction containing specific action guidance according to the risk assessment result; an early warning cooperative execution module for executing the early warning decision instruction and synchronously distributing early warning information to a train operation terminal and an operation terminal.

2. The multi-source data fusion based train regional perception and active protection system according to claim 1, wherein, The acquisition of the obstacle perception data comprises: scanning an area in front of a train operation by a radar to obtain original detection data containing distance, speed and direction information; image acquisition of the area covered by the radar scanning to obtain visible light or infrared image data; fusion processing of the original detection data and the image data to identify and extract obstacle information in the track and the surrounding area to generate obstacle perception data.

3. The multi-source data fusion based train zone awareness and active protection system of claim 1, wherein, The reception of the personnel state and position data comprises: real-time acquisition of identity information, vital sign state data and position data based on a Beidou satellite navigation system by an intelligent safety terminal carried by an operation personnel; uploading the identity information, state data and position data to the system through a wireless communication network; receiving and analyzing the uploaded data as personnel state and position data.

4. The multi-source data fusion based train regional perception and active protection system of claim 1, wherein, The generation and management of the electronic fence comprises: analyzing accessed operation plan data to extract location information and a time window of an operation section; defining a corresponding graphical area on an electronic map of a railway line based on the location information to generate an electronic fence; associating the time window with the electronic fence to activate, continuously monitor and update the state of the electronic fence.

5. The multi-source data fusion based train regional perception and active protection system of claim 1, wherein, The fusion processing comprises: spatial and temporal alignment of the obstacle perception data and the personnel state and position data to a space-time reference defined by the electronic map; associating a target in the obstacle perception data with an individual in the personnel state and position data as a spatial constraint condition; based on the association result, integrating the train, the personnel, the electronic fence and the identified risk area to generate global dynamic situation information.

6. The multi-source data fused train area awareness and active protection system of claim 1, wherein, The calculation of the time to collision (TTC) between the train and the operation personnel comprises: real-time acquisition of train real-time position and speed information and operation personnel real-time position information contained in the global dynamic situation information; calculation of the relative distance between the train and the operation personnel based on the real-time position information of the two; calculation of the approaching speed of the train relative to the operation personnel based on the speed information of the train and the movement information of the operation personnel; The collision time TTC is calculated according to the relative distance and the approaching speed.

7. The multi-source data fused train area awareness and active protection system of claim 1, wherein, The dynamic risk threshold model comprises: The TTC threshold range corresponding to different risk levels is dynamically determined according to the train type, running speed and line condition; The mapping relationship from the TTC value to the risk level is established based on the TTC threshold range; In real-time risk assessment, the corresponding risk level is output according to the calculated TTC value and the mapping relationship.

8. The multi-source data fused train area awareness and active protection system of claim 1, wherein, The generation of the hierarchical early warning decision instruction comprises: According to the risk level in the risk assessment result, the corresponding instruction template in the preset early warning strategy library is matched; Combined with the personnel position and line environment in the global dynamic situation information, an action guide containing specific avoidance direction, recommended operation intensity and estimated remaining time is generated; The action guide and the corresponding early warning level are encapsulated to generate a hierarchical early warning decision instruction.

9. The multi-source data fused train area awareness and active protection system of claim 1, wherein, The distribution of the early warning information comprises: The early warning decision instruction is converted into a communication protocol and data format suitable for the train operation terminal and the operation terminal respectively, and a to-be-distributed early warning data packet is generated; Through the railway special wireless communication network, the to-be-distributed early warning data packet is synchronously sent to the corresponding train operation terminal and operation terminal; The terminal triggers the early warning prompt to be output through at least two channels of the visual display interface, the audible alarm device and the tactile vibration device according to the early warning data packet.

10. A method for train regional perception and active protection of multi-source data fusion, characterized in that, The train regional perception and active protection system for the multi-source data fusion of any one of claims 1-9, the method comprising: Obtain obstacle perception data of the area in front of the train running, and receive personnel state and position data reported by the field operation terminal; Provide positioning data of the train and the personnel, a railway line electronic map, and generate and manage an electronic fence on the electronic map according to the operation plan; Fuse the obstacle perception data, the personnel state and position data, the electronic map and the electronic fence to generate global dynamic situation information; Based on the global dynamic situation information, calculate the collision time TTC between the train and the operation personnel, and output a risk assessment result according to the TTC and a preset dynamic risk threshold model; According to the risk assessment result, generate a hierarchical early warning decision instruction containing specific action guide; Execute the early warning decision instruction, and synchronously distribute early warning information to the train operation terminal and the operation terminal.

Images