A method for matching complex road segment beidou trajectory with road network topology

Abstract

This application belongs to the field of BeiDou high-precision positioning technology for smart highways, and specifically discloses a method for matching BeiDou trajectories with road network topology in complex road sections. The method includes: acquiring and normalizing trajectory and road network data; removing outliers and drift points from the BeiDou trajectory data; performing time-series alignment based on the road network topology data; converting the time-aligned trajectory coordinates to planar coordinates; using a three-level road network grid index for hierarchical retrieval at the segment, lane, tunnel, or interchange levels; calculating the distance from the trajectory point to the centerline of the road network within the grid; obtaining the road network topology-constrained positioning result based on the minimum distance principle; extracting the road network curvature and trajectory heading angle to construct curvature-heading coupling feature values; segmenting the tunnel and setting adaptive fusion weights for the inertial positioning result and the road network topology-constrained positioning result; and fusing the curvature-heading coupling feature values ​​to obtain the final matched positioning coordinates. This application can effectively improve the matching accuracy of BeiDou trajectories and road network topology in complex road sections.

Description

Technical Field

[0001] This application belongs to the field of BeiDou high-precision positioning technology for smart highways, and more specifically, relates to a road network topology matching method for BeiDou trajectories in complex road sections. Background Technology

[0002] Long highway tunnels (tunnels longer than 3000 meters) suffer from problems such as BeiDou signal blockage, large positioning blind spots, and severe trajectory drift. Multi-level interchanges are characterized by dense road network topology, overlapping lanes, and drastic changes in heading. Existing trajectory matching methods only employ single spatial matching or fixed inertial compensation, which have shortcomings in the above-mentioned complex road sections, such as low matching accuracy, incomplete drift point elimination, and distortion of blind spot trajectory completion. The accurate matching rate is less than 90%, which cannot meet the industry's needs for real-time accurate monitoring and route tracing of operating vehicles.

[0003] Therefore, improving the matching accuracy between BeiDou trajectories and road network topology in complex road sections is an urgent problem to be solved. Summary of the Invention

[0004] In view of the shortcomings of the existing technology, the purpose of this application is to provide a road network topology matching method for Beidou trajectories in complex road sections, which can effectively improve the matching accuracy between Beidou trajectories and road network topology in complex road sections.

[0005] To achieve the above objectives, in a first aspect, this application provides a road network topology matching method for BeiDou trajectories in complex road sections, comprising the following steps: S10: Acquire BeiDou trajectory data and road network topology data and perform coordinate normalization; remove outliers and drift points from the normalized BeiDou trajectory data to obtain purified trajectory points; S20, based on the normalized road network topology data, performs time-series standardization and time dimension alignment on the purified trajectory points to achieve time synchronization between Beidou trajectory data and road network topology data; S30: Convert the time-aligned trajectory point coordinates into Cartesian coordinates; use a pre-built three-level road network grid index, and perform hierarchical retrieval according to the rules of three levels: road segment level, lane level, and tunnel or interchange level; calculate the distance from the trajectory point to the center line of the road network in each grid; determine the optimal matching grid based on the minimum distance principle; and use the coordinates of the center line of the road network corresponding to the optimal matching grid as the road network topology constraint positioning result. S40, extract the road network curvature and trajectory heading angle to construct curvature-heading coupling feature values; divide the tunnel into multiple segments and set adaptive fusion weights for the inertial positioning results and the road network topology constraint positioning results within each segment; based on the curvature-heading coupling feature values, fuse the inertial positioning results and the road network topology constraint positioning results according to the fusion weights corresponding to the segment where the current trajectory point is located to obtain the final matching positioning coordinates.

[0006] As a further preferred embodiment, in step S10, the outlier removal specifically includes: Constructing velocity anomaly factors , ,in The instantaneous velocity of the trajectory point is measured. The legally mandated speed limit for this road section; Constructing road network distance anomaly factors , ,in The vertical distance from the trajectory point to the centerline of the road network. The maximum permissible road network deviation distance is preset; Constructing a two-dimensional joint outlier discriminant index , ,in For the velocity anomaly factor weighting coefficient, For the road network distance anomaly factor weight coefficient, and + =1; Set outlier threshold ,when ≥ When it is identified as an outlier and removed, < The data is retained as valid trajectory points to complete the purification of BeiDou trajectory data.

[0007] As a further preferred embodiment, step S10, the removal of drift points specifically includes: Calculate the score of satellite observations , ,in For the number of effective satellites, The preset satellite number threshold; Computation speed compliance score , ,in The instantaneous velocity of the trajectory point is measured. The legally mandated speed limit for this road section; Calculate the road network topology distance score , ,in The vertical distance from the trajectory point to the centerline of the road network. The preset maximum allowable distance; Calculate the heading angle stability score , ,in This represents the change in heading between adjacent points. The preset maximum allowable angle of change; Will , , , Through weighted formula Calculate the total confidence score of trajectory points Trajectory points with S < 0.8 are identified as drift points and removed, while high-scoring valid points with S ≥ 0.8 are retained.

[0008] As a further preferred embodiment, the method for determining the heading change Δθ between adjacent points is as follows: calculate the original difference in heading angles. ,in Let be the driving direction angle of the i-th trajectory point. Let θ be the heading angle of the (i-1)th trajectory point; then, Δθ is obtained by correcting it according to the following rule: if Then Δθ = ;like Then Δθ = .

[0009] As a further preferred embodiment, in step S20, the time series standardization and time dimension alignment specifically include: based on the original BeiDou timestamp, constructing a 1-second fixed interval standard time axis, using linear interpolation to fill in the missing time points for the original trajectory points with non-equal intervals, and generating an equal interval standardized trajectory; binding the road network topology data to the 1-second fixed interval standard time series to complete the time dimension association between the road network and the trajectory.

[0010] As a further preferred embodiment, in step S30, the three-level road network grid index is constructed as follows: using the CGCS2000 coordinate system as the spatial reference, the main line, ramps, and connecting lines of the expressway are divided into first-level grids at the road segment level, with the grid boundaries forcibly aligned with the road network topology nodes, and a unique code using the road segment number plus the grid sequence number is adopted; within the road segment level grid, second-level sub-grids at the lane level are divided according to the overtaking lane, driving lane, and emergency lane; for tunnels or multi-level interchanges with a length greater than 3000 meters, a third-level grid at the tunnel or interchange level is generated with densification, and signal blind spots or overlapping areas of the road network are marked; a hierarchical mapping relationship and a hash index table are constructed.

[0011] As a further preferred embodiment, in step S40, the curvature-heading coupling characteristic value The construction formula is ,in For real-time road network curvature, The reference curvature of the road segment For the trajectory heading angle, The standard heading angle for the road network. This is the curvature weighting coefficient. This is the heading angle weighting coefficient; The tunnel is divided into multiple sections as follows: the entrance section is 0-200 meters, the middle section is 200 meters to the tunnel end minus 200 meters, and the exit section is the tunnel end minus 200 meters to the end. The rule for setting the adaptive fusion weights is as follows: Entry segment =0.6、 =0.4, middle section =0.8、 =0.2, Exit Section =0.4、 =0.6, where As the weight of the inertial positioning results, Weights are assigned to the location results based on road network topology constraints, and + =1; Based on the curvature-heading coupling eigenvalue The adaptive fusion weights are dynamically fine-tuned; The specific formula for the fusion is as follows: ,in To ultimately fuse the positioning coordinates, This is the result of inertial positioning. The location results are based on road network topology constraints.

[0012] As a further preferred embodiment, a trajectory coherence backtracking correction step is also included: after obtaining the road network topology constraint positioning results and before fusion, a trajectory coherence verification index is constructed for five consecutive trajectory points. , ,in For the first i Matching coordinates of trajectory points The network topology reference coordinates are the coordinates of the i-th trajectory point; a preset continuity judgment threshold is used. The matching deviation exceeded the threshold in 5 consecutive points. Number of outliers ,like If the value is ≥ 2, then backtrack to the 10 valid trajectory points before the current anomaly point and construct the backtracking correction coordinate fusion formula. ,in For the first j The corrected coordinates of the outliers Let K be the coordinates of the kth valid trajectory point. The temporal adaptive weights are used for the k-th valid trajectory point; the original abnormal matching points are replaced with the corrected coordinates.

[0013] As a further preferred option, a multi-source fusion calibration iteration step is also included, specifically: After obtaining the final matching and positioning coordinates, a multi-source weighted fusion positioning formula is constructed. ,in , , For adaptive weights and + + =1, For multi-source fusion positioning coordinates, The original positioning coordinates for BeiDou. Provide the coordinates for radar vision positioning. Provide the reference positioning coordinates for the ETC gantry; Constructing a multi-source error verification formula ,in The total verification error of multi-source fusion =∣ | represents the lane matching error for radar vision. =∣ | This indicates an error in the absolute location of the ETC (Electronic Toll Collection) system. when ≤1m and Points ≤5m are marked as precise matching points and fed back to the third-level road network grid index. The formula is then iteratively optimized through grid parameters. Update the grid index parameters, where , Here, λ represents the grid index parameters before and after the iteration, and λ is the iteration step size coefficient. This represents the global average matching error.

[0014] Secondly, this application provides a road network topology matching system for BeiDou trajectories of complex road sections, including a processor and a memory. The memory stores a computer program, and when the processor executes the computer program, it implements the road network topology matching method for BeiDou trajectories of complex road sections as described in any one of the above.

[0015] This application has the following beneficial effects: (1) In the prior art, the original trajectory data of vehicle-mounted Beidou is easily affected by abnormal vehicle driving, satellite obstruction and terminal operating condition fluctuations. There are generally outlier data such as speeding anomalies, large geographical location deviations and positioning drift. At the same time, the sampling frequency of vehicle-mounted Beidou terminal and road network basic data are not consistent and there is a time delay in communication transmission. The two types of data naturally have spatiotemporal misalignment problems, which directly cause the subsequent trajectory matching benchmark to be inaccurate and the reliability of the underlying data to be insufficient.

[0016] This application leverages data purification and temporal alignment technologies to improve the quality of raw data. The solution establishes a dual-dimensional outlier removal model based on speed and road network distance, accurately identifying and eliminating abnormal trajectory points such as those exceeding speed limits or having excessive positional deviations. Simultaneously, a four-dimensional trajectory credibility weighted evaluation system is established, incorporating satellite observation quantity, speed compliance, road network topology distance, and heading angle stability. Low-scoring trajectory points are identified as data drift points and eliminated. This dual data purification mechanism ensures the validity and reliability of trajectory data from the source. Furthermore, through timestamp normalization, equal-interval resampling, temporal interpolation, and data calibration, the spatiotemporal offset problem caused by asynchronous sampling and data transmission delays between the BeiDou vehicle-mounted terminal and the road network baseline data is addressed. This establishes a unified temporal benchmark for trajectory matching, avoiding matching deviations caused by temporal discrepancies.

[0017] (2) For scenarios with dense and complex road networks such as extra-long tunnels and multi-level interchanges, the existing conventional single-level road network index has a large retrieval range and is prone to road mismatch at intersections. Direct calculation of spherical latitude and longitude introduces inherent spherical geometric errors. Inertial positioning generates drift errors as mileage accumulates. Furthermore, there is a lack of targeted positioning compensation schemes for the complete blind zone of tunnel satellite signals. This easily leads to the drawbacks of trajectory deviating from the real road network and trajectory completion distortion in blind zones.

[0018] This application employs a hierarchical retrieval combined with geometric constraints to improve trajectory matching accuracy. A three-tiered road network grid index system is established: segment-level, lane-level, and tunnel and interchange-level special road segments. A coarse-to-fine hierarchical retrieval strategy effectively narrows the spatial retrieval range, resolving retrieval disorder and road mismatch issues caused by dense road networks and overlapping road segments. Gaussian projection is used to convert latitude and longitude to Cartesian coordinates, eliminating geometric errors introduced by spherical coordinates. The optimal matching grid is selected using minimum Euclidean distance, achieving high-precision lane-level matching. Curvature-heading coupled feature quantities are constructed, using road network curvature and heading angle as geometric constraints to ensure the trajectory conforms to the actual road network shape, suppressing accumulated errors from inertial positioning. For tunnel satellite signal blind spots, the tunnel is divided into three sections: entrance, middle section, and exit. A segmented adaptive weight fusion of inertial positioning and road network topology constraint positioning results is adopted: when signal obstruction is severe in the middle section of the tunnel, the inertial positioning weight is increased to ensure positioning continuity; after signal recovery at the tunnel entrance and exit, the road network constraint weight is increased to correct trajectory deviation, thoroughly improving the problems of trajectory drift and data completion distortion in blind spots.

[0019] (3) Traditional trajectory matching often adopts an open-loop calculation mode. Single-point matching errors are easily propagated and spread along the time axis, resulting in continuous mismatch of the entire trajectory. Moreover, relying solely on a single Beidou positioning data source cannot utilize ETC and multi-source measured information from radar vision to calibrate positioning deviations. The road network matching benchmark cannot be dynamically iterated and optimized, resulting in a trajectory matching rate of less than 90% for complex road sections such as extra-long tunnels and multi-level interchanges.

[0020] This application utilizes backtracking correction and multi-source data closed-loop calibration to block error propagation, achieving full-process iterative optimization and significantly improving the matching success rate of complex road sections. A trajectory continuity verification and backtracking correction mechanism is established, calculating continuity indicators for continuous trajectory points. When the number of abnormal points exceeds a preset threshold, the system backtracks to the 10 valid trajectory points preceding the current abnormal point, and recalibrates the coordinates using time-adaptive weights, effectively reducing the impact of single-point matching errors on the entire trajectory and preventing error propagation. This solution integrates three data sources: BeiDou positioning, radar vision sensing, and ETC gantry data collection, designing a multi-source weighted fusion positioning model and error verification model. When both the radar vision lane matching error and the ETC absolute position error are below preset thresholds, the point is identified as a high-precision matching point and transmitted back to the three-level road network grid indexing system. By continuously updating the matching benchmark based on the grid parameter iterative optimization algorithm, a closed-loop system is formed that includes data purification, spatiotemporal alignment, grid retrieval, spatial matching, geometric constraints, backtracking correction, and multi-source iteration. Ultimately, the trajectory matching rate of complex road sections is increased from below 90% to 98.1%. Attached Figure Description

[0021] Figure 1 This is a flowchart of the method provided in the embodiments of this application; Figure 2 This is a matching accuracy response curve diagram for a particularly long tunnel scenario provided in an embodiment of this application; Figure 3 This is a horizontal comparison bar chart of the prior art and the method provided in this application embodiment on the accuracy matching rate index; Figure 4 This is a horizontal comparison bar chart of the prior art and the present method in terms of outlier removal rate provided in the embodiments of this application; Figure 5 This is a horizontal comparison bar chart of the prior art and the method provided in this application embodiment on the trajectory continuity pass rate index. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0023] like Figure 1 As shown, this embodiment provides a road network topology matching method for BeiDou trajectories in complex road sections, including the following steps: S1, Data Preprocessing.

[0024] The CGCS2000 national geodetic coordinate system was uniformly adopted as the spatial reference to complete the coordinate normalization of BeiDou trajectory and road network topology data, and to eliminate coordinate system deviations of multi-source data.

[0025] It should be noted that the BeiDou trajectory data is a sequence of vehicle dynamic positioning points collected over time by the vehicle-mounted BeiDou terminal, and the stored fields are: CGCS2000 latitude and longitude coordinates, instantaneous vehicle speed. Heading angle Number of effective observation satellites n Data collection timestamp. Road network topology data is static geographic benchmark data describing the actual connections between roads, lanes, tunnels, and road segments. Stored fields include: road network centerline coordinates, and legal speed limits for road segments. Road curvature C Road network standard heading Road segment spatial topology connections and lane layering attributes.

[0026] A dual-dimensional outlier removal model based on speed and road network topology distance is constructed to accurately identify and remove speeding and out-of-distance abnormal trajectory points, as detailed below: Velocity anomaly factor: The larger the value, the more significant the deviation from the speed limit. The speed anomaly factor characterizes the degree to which the speed of the trajectory point deviates from the speed limit of the road section; The measured instantaneous velocity at the trajectory point (unit: km / h); The legal speed limit for this road section (unit: km / h).

[0027] Road network distance anomaly factors: A higher value indicates a more severe deviation of the trajectory from the road network. Among these, This is the road network distance anomaly factor, representing the degree to which trajectory points deviate from the road network; The vertical distance from the trajectory point to the center line of the road network (unit: m) is calculated approximately using latitude and longitude on a spherical surface, and outliers can be coarsely removed. The maximum permissible road network deviation distance is preset (unit: m).

[0028] Two-dimensional joint outlier discriminant index: Set outlier detection threshold ,when ≥ When, it is determined to be an invalid trajectory point (outlier point) and directly removed; when < At that time, points deemed valid are retained, completing the original data purification process. It is a two-dimensional joint outlier discriminant index. The weighting coefficient of the velocity anomaly factor (0 < <1), The weighting coefficient of the road network distance anomaly factor (0 < <1), + =1.

[0029] S2, trajectory credibility weighted score.

[0030] This step constructs a four-dimensional trajectory credibility scoring system based on satellite observation data, speed compliance, road network topology distance, and heading angle stability. Through sub-item scoring formulas and a weighted fusion model, drift trajectory points are accurately identified and eliminated to ensure the purity and reliability of subsequent matching data.

[0031] Satellite observation score : based on the number of effective satellites Based on this, a threshold for the number of satellites is set. The scoring formula is: The more satellites a score has, the higher the score; if the score is below the threshold, the score will decrease significantly.

[0032] Speed ​​compliance score Comparison with actual measured speed Speed ​​limits for the road section The compliance score formula is: The closer the speed is to the speed limit, the higher the score; speeding or going too slow will both result in a deduction of points.

[0033] Road network topology distance score Calculate the perpendicular distance from the trajectory point to the road network. Set the maximum allowable distance The distance score formula is: The closer you are to the road network, the higher your score; if you are too far away, you get 0 points.

[0034] Heading angle stability score Calculate the change in heading between adjacent points The details are as follows: 1) Calculate the original difference in heading angle

[0035] Let be the driving direction angle of the i-th trajectory point (e.g., 90° if the car is facing east, and 0° / 360° if it is facing north). The heading angle of the point preceding it; This is the original difference obtained by directly subtracting the two.

[0036] 2) Correct the ring angle and solve for the actual change in heading. The heading is a circular angle, and the minimum effective turning angle between two points will not exceed [a certain value]. There are two scenarios to consider:

[0037] This is the final change in heading between adjacent points, with a range of 0° ≤ ≤180°.

[0038] Set the maximum allowable angle of change The stable score formula is: The more stable the course, the higher the score; sudden changes will result in a significant deduction of points.

[0039] Substitute the four scores into the weighted formula: .

[0040] Filtered by rules: Total credibility score Trajectory points with a score <0.8 are considered drift points and are directly removed; only high-scoring valid points are retained for subsequent processes.

[0041] S3, time-series standardization and time dimension alignment.

[0042] This step involves unified time reference processing based on the effective BeiDou trajectory dataset after S1 coordinate normalization and S2 filtering and purification (the spatial reference has already been completed in S1 using CGCS2000), constructing a unified time series system for the entire dataset: Based on the original BeiDou timestamps, a standard time axis with a fixed interval of 1 second is constructed. For the original trajectories with uneven sampling density and disordered timestamps, linear interpolation is used to fill in the missing time points and generate standardized trajectories with equal intervals, so as to realize the continuous temporal sequence of vehicle trajectories. By binding static road network topology data to the aforementioned standard 1-second time series, the temporal dimension association between the road network and the trajectory is completed; By normalizing timestamps and interpolating, the timing offset error caused by asynchronous sampling and transmission delay in BeiDou terminal and road network data is eliminated.

[0043] S4, three-level grid index construction.

[0044] Using the CGCS2000 coordinate system as the spatial reference, a three-level road network grid index is constructed, hierarchically: road segment level → lane level → tunnel / interchange level, to achieve refined retrieval and management of complex road networks. First, road segment-level grids are divided, with independent road segments such as highway mainlines, ramps, and connecting lines classified as first-level grids and assigned unique road segment codes to quickly locate the large interval to which the trajectory belongs. Then, lane-level grids are further subdivided within the road segment-level grids, with second-level sub-grids divided according to overtaking lanes, driving lanes, and emergency lanes to match lane-level positioning requirements. Finally, for complex scenarios such as extra-long tunnels and multi-level interchanges, a third-level tunnel / interchange grid is generated with increased density, reducing grid size and marking signal blind spots and overlapping areas of the road network. A hierarchical mapping relationship and hash index table are constructed. During trajectory matching, a "coarse-to-fine" retrieval order is used: first, the road segment-level grid is located, then the lane-level grid is searched, and finally, the special scenario-level grid is located. This significantly reduces the spatial retrieval range, improves the efficiency of trajectory matching for complex road segments, and provides efficient index support for subsequent spatial matching and inertial compensation.

[0045] Specifically, the granularity quantification rules for road segment-level grid partitioning are as follows: For road segment-level (Level 1) grids, clear quantification standards, boundary definitions, and coding rules are established to resolve the issue of ambiguous partitioning granularity. The core partitioning principle is based on the physical road segment attributes and mileage segmentation of highways, ensuring grid independence, uniqueness, and retrieval efficiency. Specific quantification rules are as follows: In terms of length granularity, the mainline is partitioned into fixed lengths of 500m ± 50m, ramps or connecting lines into 100m ± 20m, and extra-long tunnel sections into 200m ± 30m. In terms of boundary definition, grid boundaries are forcibly aligned with road network topology nodes (interchange nodes, tunnel entrances, toll stations, ramp divergence or merging points), and do not cross physical road segment boundaries. In terms of attribute partitioning, five types of physical road segments are independently partitioned: mainline, left ramp, right ramp, emergency connecting line, and tunnel body, each type being an independent road segment-level grid. The coding rules use a unique code of road segment number plus grid sequence number, for example, GXX-mainline-001, supporting fast retrieval via hash index. Regarding spatial reference constraints, all road segment-level grids are generated based on the CGCS2000 coordinate system, with a planar projection error ≤0.5m.

[0046] S5, refined spatial matching.

[0047] The latitude and longitude coordinates of the BeiDou trajectory CGCS2000 are converted to Gaussian plane rectangular coordinates through Gaussian projection forward calculation to eliminate spherical geometric errors. Based on the constructed three-level road network grid index, according to the hierarchical retrieval rules of "segment level → lane level → tunnel / interchange level", the Euclidean distance from the trajectory point to the center line of the road network in each grid is calculated, and the optimal matching grid is determined by the principle of minimum distance. Taking the main line segment level grid as an example, the grid is divided into three types of lane-level sub-grids: overtaking lane, driving lane, and emergency lane. Each sub-grid corresponds to an independent road network center line. The plane rectangular coordinates of the trajectory point are substituted into the Euclidean distance formula to calculate the vertical distance from it to the center line of the road network in each sub-grid. The lane-level sub-grid corresponding to the minimum distance is selected as the optimal matching grid for that trajectory point. For complex scenarios such as tunnels and interchanges, the same rules are used to select the optimal matching result from the encrypted tunnel / interchange level grid to achieve accurate spatial matching of the trajectory point at the lane level. By comparing point-by-point planar coordinates and searching the grid, the road segment and lane to which the trajectory point belongs can be quickly located, and the initial refined spatial matching can be completed, providing accurate initial spatial positioning basis for subsequent road network curvature-heading angle coupling matching and tunnel inertia compensation.

[0048] S6, Road network curvature-heading angle coupling matching and tunnel segment inertia compensation.

[0049] This step achieves accurate trajectory completion in BeiDou signal blind spots by adaptively fusing the road network curvature-heading angle coupling features with the tunnel three-segment IMU inertial positioning results and road network weights. First, the road network segment curvature and trajectory heading angle are extracted, a curvature-heading coupling feature formula is constructed, and the geometric fit between the trajectory and the road network is determined. ,in For curvature-heading coupling eigenvalues, For real-time road network curvature, The reference curvature of the road segment For the trajectory heading angle, The standard heading angle for the road network. This is the curvature weighting coefficient. This is the heading angle weighting coefficient.

[0050] The tunnel is divided into an entrance section (0-200m), a middle section (200m to the tunnel end minus 200m), and an exit section (the tunnel end minus 200m to the end). An adaptive weight fusion formula is constructed based on the segmented scenario: ,in For final fusion of location coordinates; This is the result of IMU inertial positioning. The location results are based on road network topology constraints. , For IMU and road network adaptive weights.

[0051] The results are as follows: The vehicle-mounted terminal integrates an accelerometer and a gyroscope IMU module, and the vehicle collects raw angular velocity and acceleration data at high frequency throughout the entire journey; based on the 1-second standard timing unified by S3, strapdown inertial calculations are performed, calculating the vehicle's displacement and heading second by second, and outputting the raw inertial position; all calculated coordinates are uniformly converted to the CGCS2000 coordinate system specified by S1 to obtain the inertial positioning results. . The results are as follows: Based on the preprocessed and purified road network vector data of S1 and the three-level grid index built in S4; after Gaussian projection to a plane and minimum Euclidean distance lane matching in S5, the trajectory point constraints are mapped to the corresponding road network centerline, and the bound coordinates of this road network centerline are... .

[0052] The system identifies tunnel segment locations in real time and automatically switches weights: entrance segment =0.6、 =0.4; middle section =0.8、 =0.2; Exit section =0.4、 =0.6, and + =1.

[0053] Based on curvature-heading coupling characteristics, road network segments are screened and trajectory morphology is verified. The degree of trajectory fit with the road network is determined based on the magnitude of the coupling feature value. Combining the tunnel entrance / middle section / exit segment division, a baseline weight is first determined and then dynamically fine-tuned to obtain segmented adaptive weights. The IMU inertial positioning results and the road network topology constraint positioning results are weighted and fused together to force the trajectory to fit the actual road network direction, effectively correcting the trajectory drift problem in blind spots. In the middle section, the BeiDou signal is severely blocked, so the IMU weight is increased to ensure positioning continuity; as the signal at the entrance and exit gradually recovers, the accuracy of the trajectory calibration is improved by the road network weight. Through curvature-heading coupling characteristics and segmented adaptive weights, the problems of trajectory drift and completion distortion in tunnel blind spots are completely solved, achieving accurate trajectory matching across the entire segment.

[0054] S7, trajectory continuity backtracking correction.

[0055] This embodiment proposes a continuous trajectory coherence verification and backtracking correction mechanism to solve single-point matching errors: This step uses a continuous trajectory consistency verification and fixed-point backtracking calibration mechanism to accurately identify and correct single-point matching errors, block error propagation, and ensure the continuity and reliability of the trajectory.

[0056] First, an original trajectory coherence verification index is constructed for 5 consecutive trajectory points to quantify the degree of matching deviation: ,in As a consistency check index, Match coordinates for the i-th trajectory point. Let be the road network topology reference coordinates corresponding to the i-th trajectory point.

[0057] Preset coherence determination threshold Count the number of outliers with matching deviations exceeding a threshold in a consecutive set of 5 points. If Nerr≥2, it is determined that there is a local matching anomaly in the trajectory, and the backtracking correction process is immediately initiated.

[0058] Tracing back to the 10 valid trajectory points preceding the current anomaly, a backtracking correction coordinate fusion formula is constructed to recalibrate the spatiotemporal reference and matching parameters: ,in The corrected coordinates of the j-th outlier point. Let K be the coordinates of the kth valid trajectory point. The time-adaptive weights are for the kth valid trajectory point (the weights for the nearest time point are higher to ensure that the correction fits the real trajectory).

[0059] After correction, the original abnormal matching points are replaced with calibration coordinates and reinstated into the matching statistics, completely eliminating the impact of single-point errors on the overall trajectory accuracy and resolving the problem of single-point matching errors in complex road sections.

[0060] S8, multi-source fusion calibration iteration.

[0061] This step employs a triple data fusion calibration using BeiDou, Raide-Vision sensing, and ETC gantry. Through multi-source fusion formulas and closed-loop iterative mechanisms, it continuously optimizes the three-level grid index and matching benchmark, comprehensively improving the trajectory matching accuracy of complex road sections.

[0062] First, a multi-source weighted fusion localization formula is constructed to fuse localization results from three types of data sources: ,in , , For adaptive weights, satisfying + + =1; For multi-source fusion positioning coordinates; The original BeiDou positioning coordinates (CGCS2000 latitude and longitude) are obtained directly by receiving satellite signals through the vehicle-mounted BeiDou terminal. The coordinates for radar-based perception and positioning (CGCS2000 latitude and longitude) are generated by the roadside radar-based integrated radar-visual device through the fusion of radar ranging and visual recognition. The reference coordinates (CGCS2000 latitude and longitude) for the ETC gantry are static reference coordinates obtained through professional surveying.

[0063] A multi-source error verification formula is constructed based on fused coordinates to quantify matching deviation: ,in =∣ | represents the lane matching error for radar vision. =∣ | represents the absolute position error of the ETC.

[0064] Determined according to calibration rules: ≤1m and When the distance is ≤5m, mark it as a precise matching point.

[0065] The precisely matched dataset is fed back to the three-level grid index module to construct an iterative optimization formula for grid parameters: ,in , Here, λ represents the grid index parameters before and after the iteration, and λ is the iteration step size coefficient. This represents the global average matching error.

[0066] The beneficial effects of this embodiment are as follows: (1) Continuous support for the trajectory in the signal blind zone: In response to the complete or partial obstruction of the Beidou signal in the tunnel, this embodiment constructs a tunnel three-segment IMU and road network weight adaptive fusion mechanism in S6. The IMU and road network weight are automatically switched at the entrance, middle section and exit. When the signal obstruction in the middle section is severe, the IMU weight is increased to ensure the continuity of positioning. When the signal at the entrance and exit is restored, the calibration accuracy of the road network weight is increased, thus avoiding the interruption of the trajectory in the blind zone in principle; (2) Drift suppression and error convergence: Combined with the curvature-heading coupling matching in S6, the road network curvature and heading angle are used as geometric constraints to force the trajectory to conform to the real road network direction, thus blocking the accumulation error of inertial positioning in principle; in conjunction with the trajectory continuity backtracking correction in S7, the error is identified and Correcting local drift, achieving a rapid reduction of drift error in the tunnel from 1.2m to 0.2m; (3) Blind spot completion and restoration: abandoning the traditional fixed inertial compensation, adopting spatiotemporal alignment, road network topology constraints, and segmented weight fusion, the completion trajectory strictly fits the tunnel alignment, eliminating completion distortion in principle; (4) Accurate retrieval of complex road networks: constructing a three-level grid index at the road segment level, lane level, and tunnel or interchange level in S4, first coarse and then fine to narrow the retrieval range, solving the retrieval chaos and matching error caused by dense road networks and overlapping intersections in principle; (5) Accurate matching at the lane level: in the refined spatial matching of S5, the coordinates are converted to plane rectangular coordinates to eliminate spherical errors, locking the optimal lane grid with the minimum Euclidean distance, from the original Theoretically solve the lane mismatch caused by lane intersection overlap; (6) Adaptation to drastic changes in heading: add heading angle stability score in S2 and curvature-heading coupling feature in S6 to quantify and score the sudden change in heading and impose geometric constraints, so as to adapt to the scenario of drastic changes in heading at interchanges in principle and avoid matching failure caused by heading jump; (7) Complete elimination of drift points: abandon the traditional single spatial filtering, construct speed-road network distance dual-dimensional outlier elimination in S1, and construct four-dimensional credibility weighted score of satellite number, speed, distance and heading in S2. The dual purification achieves 100% identification and elimination of drift points in principle; (8) Improved matching accuracy: adopt a four-layer matching mechanism of spatial matching, geometric constraints, temporal verification and multi-source calibration. , replacing the traditional single spatial matching, greatly improving the stability of matching in complex scenarios in principle; (9) Blind spot completion and fidelity: with road network topology as hard constraint, IMU as soft support, and multi-source data as calibration, the completion trajectory strictly follows the real road alignment; (10) Full process accuracy guarantee: from data purification, spatiotemporal alignment, grid indexing, spatial matching, coupling features, backtracking correction, and multi-source iteration to form a closed loop, each step provides support for accuracy improvement, and in principle, the matching rate is increased from less than 90% to 98.1%; (11) Path tracing reliability: through trajectory continuity verification and backtracking correction to block error transmission, multi-source fusion calibration provides absolute position benchmark, and in principle, path tracing is guaranteed to be 100% real and reliable.

[0067] The following is a specific implementation example of this application: Two typical and complex scenarios were selected: a 5km long tunnel and a three-level interchange. Real vehicle tests and simulation verifications were carried out.

[0068] Scenario 1: Real-world application in a 5km long tunnel. Test conditions included 90% BeiDou signal obstruction, vehicle speed of 60-80km / h, and the vehicle being equipped with BeiDou, IMU, Raiden, and ETC devices, undergoing three consecutive passages. Regarding the core response curve, the matching accuracy response curve (e.g.) Figure 2 As shown in the figure: from the entrance to the middle section to the exit, the accuracy improved from 92% to 98.5% to 99%, and the drift error improved from 1.2m to 0.3m to 0.2m (the error continued to converge). In terms of performance changes, the outlier removal rate was 99.2%, the trajectory integrity rate was 99.8%, and the processing time for a single point was ≤10ms.

[0069] This embodiment compares the technology with existing technologies. Two mainstream existing technologies, the single-space matching method and the fixed inertia compensation method, are selected for comparison under the same scenario, data, and hardware. Key indicators are quantified as follows: Test conditions: Mixed and complex road section (5km tunnel + 3-level interchange), 2000 trajectory samples, 1Hz sampling rate, CGCS2000 unified benchmark.

[0070] The core performance comparison data is shown in Table 1.

[0071] Table 1 Core Performance Comparison Data

[0072] like Figure 3 , Figure 4 and Figure 5 The bar chart shown illustrates a horizontal comparison of the accuracy matching rate, outlier removal rate, and consistency pass rate. It can be seen that the accuracy matching rate of the method of this invention exceeds 98%, and its adaptability to complex road sections, real-time performance, and stability are significantly better than existing technologies.

[0073] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A road network topology matching method for BeiDou trajectories in complex road sections, characterized in that, Includes the following steps: S10: Acquire BeiDou trajectory data and road network topology data and perform coordinate normalization; remove outliers and drift points from the normalized BeiDou trajectory data to obtain purified trajectory points; S20, based on the normalized road network topology data, performs time-series standardization and time dimension alignment on the purified trajectory points to achieve time synchronization between Beidou trajectory data and road network topology data; S30: Convert the time-aligned trajectory point coordinates into Cartesian coordinates; use a pre-built three-level road network grid index, and perform hierarchical retrieval according to the rules of three levels: road segment level, lane level, and tunnel or interchange level; calculate the distance from the trajectory point to the center line of the road network in each grid; determine the optimal matching grid based on the minimum distance principle; and use the coordinates of the center line of the road network corresponding to the optimal matching grid as the road network topology constraint positioning result. S40, extract road network curvature and trajectory heading angle, and construct curvature-heading coupled feature values; The tunnel is divided into multiple segments, and an adaptive fusion weight is set between the inertial positioning results and the road network topology constraint positioning results in each segment. Based on the curvature-heading coupling feature value, the inertial positioning result and the road network topology constraint positioning result are fused according to the fusion weight corresponding to the segment where the current trajectory point is located, to obtain the final matching positioning coordinates.

2. The road network topology matching method for complex road sections using BeiDou trajectories as described in claim 1, characterized in that, In step S10, the outlier removal specifically includes: Constructing velocity anomaly factors , ,in The instantaneous velocity of the trajectory point is measured. The legally mandated speed limit for this road section; Constructing road network distance anomaly factors , ,in The vertical distance from the trajectory point to the centerline of the road network. The maximum permissible road network deviation distance is preset; Constructing a two-dimensional joint outlier discriminant index , ,in For the velocity anomaly factor weighting coefficient, For the road network distance anomaly factor weight coefficient, and + =1; Set outlier threshold ,when ≥ When it is identified as an outlier and removed, < The data is retained as valid trajectory points to complete the purification of BeiDou trajectory data.

3. The road network topology matching method for complex road sections using BeiDou trajectories as described in claim 1, characterized in that, In step S10, the drift point removal specifically includes: Calculate the score of satellite observations , ,in For the number of effective satellites, The preset satellite number threshold; Computation speed compliance score , ,in The instantaneous velocity of the trajectory point is measured. The legally mandated speed limit for this road section; Calculate the road network topology distance score , ,in The vertical distance from the trajectory point to the centerline of the road network. The preset maximum allowable distance; Calculate the heading angle stability score , ,in This represents the change in heading between adjacent points. The preset maximum allowable angle of change; Will , , , Through weighted formula Calculate the total confidence score of trajectory points Trajectory points with S < 0.8 are identified as drift points and removed, while high-scoring valid points with S ≥ 0.8 are retained.

4. The road network topology matching method for complex road sections using BeiDou trajectories as described in claim 3, characterized in that, The method for determining the heading change Δθ between adjacent points is as follows: calculate the original difference in heading angle. ,in Let be the driving direction angle of the i-th trajectory point. Let θ be the heading angle of the (i-1)th trajectory point; then, Δθ is obtained by correcting it according to the following rule: if Then Δθ = ;like Then Δθ = .

5. The road network topology matching method for complex road sections using BeiDou trajectories as described in claim 1, characterized in that, In step S20, the time series standardization and time dimension alignment specifically include: based on the original BeiDou timestamp, constructing a 1-second fixed interval standard time axis, using linear interpolation to fill in missing time points for non-equal interval original trajectory points, generating an equal interval standardized trajectory; binding the road network topology data to the 1-second fixed interval standard time series, completing the time dimension association between the road network and the trajectory.

6. The road network topology matching method for complex road sections using BeiDou trajectories as described in claim 1, characterized in that, In step S30, the three-level road network grid index is constructed as follows: using the CGCS2000 coordinate system as the spatial reference, the main line, ramps, and connecting lines of the expressway are divided into first-level grids at the road segment level. The grid boundaries are forcibly aligned with the road network topology nodes, and a unique code of road segment number plus grid sequence number is used. Within the road segment level grid, second-level sub-grids at the lane level are divided according to overtaking lanes, driving lanes, and emergency lanes. For tunnels or multi-level interchanges with a length greater than 3000 meters, a third-level grid at the tunnel or interchange level is generated with densification, and signal blind spots or overlapping areas of the road network are marked. A hierarchical mapping relationship and a hash index table are constructed.

7. The road network topology matching method for complex road sections using BeiDou trajectories as described in claim 1, characterized in that, In step S40, the curvature-heading coupling eigenvalues The construction formula is ,in For real-time road network curvature, The reference curvature of the road segment For the trajectory heading angle, The standard heading angle for the road network. This is the curvature weighting coefficient. This is the heading angle weighting coefficient; The tunnel is divided into multiple sections as follows: the entrance section is 0-200 meters, the middle section is 200 meters to the tunnel end minus 200 meters, and the exit section is the tunnel end minus 200 meters to the end. The rule for setting the adaptive fusion weights is as follows: Entry segment =0.6、 =0.4, middle section =0.8、 =0.2, Exit Section =0.4、 =0.6, where As the weight of the inertial positioning results, Weights are assigned to the location results based on road network topology constraints, and + =1; Based on the curvature-heading coupling characteristic value The adaptive fusion weights are dynamically fine-tuned; The specific formula for the fusion is as follows: ,in To ultimately fuse the positioning coordinates, This is the result of inertial positioning. The location results are based on road network topology constraints.

8. The road network topology matching method for complex road sections using BeiDou trajectories as described in claim 1, characterized in that, It also includes a trajectory coherence backtracking correction step: after obtaining the road network topology constraint positioning results and before fusion, a trajectory coherence verification index is constructed for five consecutive trajectory points. , ,in For the first i Matching coordinates of trajectory points The network topology reference coordinates are the coordinates of the i-th trajectory point; a preset continuity judgment threshold is used. The matching deviation exceeded the threshold in 5 consecutive points. Number of outliers ,like If the value is ≥2, then backtrack to the 10 valid trajectory points before the current anomaly point and construct the backtracking correction coordinate fusion formula. ,in For the first j The corrected coordinates of the outliers Let K be the coordinates of the kth valid trajectory point. The time-adaptive weights corresponding to the kth valid trajectory point; Replace the original abnormal matching points with the corrected coordinates.

9. The road network topology matching method for complex road sections using BeiDou trajectories as described in claim 1, characterized in that, It also includes a multi-source fusion calibration iteration step, specifically including: After obtaining the final matching and positioning coordinates, a multi-source weighted fusion positioning formula is constructed. ,in , , For adaptive weights and + + =1, For multi-source fusion positioning coordinates, The original positioning coordinates for BeiDou. Provide the coordinates for radar vision positioning. Provide the reference positioning coordinates for the ETC gantry; Constructing a multi-source error verification formula ,in The total verification error of multi-source fusion =∣ | represents the lane matching error for radar vision. =∣ | represents the absolute position error of the ETC (Electronic Toll Collection) system. when ≤1m and Points ≤5m are marked as precise matching points and fed back to the third-level road network grid index. The formula is then iteratively optimized through grid parameters. Update the grid index parameters, where , Here, λ represents the grid index parameters before and after the iteration, and λ is the iteration step size coefficient. This represents the global average matching error.

10. A road network topology matching system for BeiDou trajectories in complex road sections, characterized in that, It includes a processor and a memory, wherein the memory stores a computer program, and the processor executes the computer program to implement the road network topology matching method for BeiDou trajectories of complex road sections as described in any one of claims 1 to 9.

Images