A vehicle monitoring method and system based on BeiDou positioning

Abstract

This application discloses a vehicle monitoring method and system based on BeiDou positioning. The BeiDou-based vehicle monitoring method includes: acquiring a raw BeiDou positioning dataset; acquiring a positioning dataset with quality level labels; correcting the data corresponding to ordinary and inferior labels to obtain corrected positioning data; acquiring a time-synchronized positioning dataset; obtaining the vehicle's three-dimensional coordinates based on the time-synchronized positioning dataset; matching the vehicle's position with an electronic map based on the vehicle's three-dimensional coordinates to obtain vehicle position information; obtaining the vehicle's trajectory information based on the vehicle's position information; and determining vehicle violations and identifying dangerous driving behaviors based on the vehicle's trajectory information. This application uses quality-corrected data to provide a foundation for time synchronization, high-precision timestamps to ensure the spatiotemporal consistency of position matching, and accurate trajectories to provide reliable input for behavior recognition.

Description

Technical Field

[0001] This application relates to the field of data analysis technology, specifically to a vehicle monitoring method and system based on BeiDou positioning. Background Technology

[0002] In existing technologies, vehicle monitoring methods based on BeiDou positioning mainly obtain vehicle location, speed, and time information by receiving satellite signals, and then combine this information with electronic maps and sensing devices for data processing and analysis. Regarding vehicle positioning, traditional methods typically rely on signals from a single satellite or a few satellites for positioning calculations. However, in complex environments such as urban canyons, tunnels, and dense forests, satellite signals are easily affected by obstruction and reflection, leading to decreased positioning accuracy and resulting in positioning drift or jumps.

[0003] Furthermore, in terms of positioning deviation correction in complex environments, existing technologies often rely on preset fixed correction parameters, lacking the ability to dynamically adapt to environmental characteristics. For example, in urban canyon areas, positioning deviations caused by building obstruction vary with building density and height, making it difficult for fixed correction parameters to meet the precise correction requirements of different scenarios. Simultaneously, the mechanisms for acquiring and updating environmental feature data are inadequate, relying heavily on manual collection or limited sensor data, making it difficult to reflect environmental changes in real time and comprehensively, thus affecting the effectiveness of positioning deviation correction. Summary of the Invention

[0004] The purpose of this invention is to provide a vehicle monitoring method based on BeiDou positioning to at least solve one of the above-mentioned technical problems.

[0005] One aspect of the present invention provides a vehicle monitoring method based on BeiDou positioning, the vehicle monitoring method based on BeiDou positioning comprising:

[0006] Obtain the structured raw dataset of BeiDou positioning transmitted by BeiDou satellites;

[0007] The acquired BeiDou positioning raw dataset is stratified by quality to obtain a positioning dataset with quality level labels, which include excellent labels, normal labels and poor labels.

[0008] Multi-window feature comparison and correction are initiated for the data corresponding to ordinary labels and inferior labels to obtain corrected location data;

[0009] The corrected location data is time-synchronized and corrected to obtain a time-synchronized location dataset.

[0010] Obtain the vehicle's 3D coordinates from the time-synchronized positioning dataset;

[0011] Obtain electronic maps;

[0012] Vehicle location information is obtained by matching the vehicle's three-dimensional coordinates with an electronic map.

[0013] Obtain vehicle trajectory information based on vehicle location information;

[0014] The system uses vehicle trajectory information to determine violations and identify dangerous driving behaviors.

[0015] Optionally, the step of performing multi-window feature comparison and correction on the data corresponding to ordinary labels and inferior labels to obtain corrected location data includes:

[0016] From the location dataset with quality grade labels, select the data corresponding to ordinary labels and poor-quality labels as the data to be corrected;

[0017] Add a timestamp index to each piece of data to be corrected;

[0018] Create multiple independent sliding windows and initialize and configure the parameters of each type of sliding window. Each type of sliding window has a different window duration than other types of sliding windows, and each child window within each type of sliding window has the same window duration.

[0019] The coordinate values ​​of longitude, latitude, and altitude are extracted from the data to be corrected using each type of sliding window.

[0020] Calculate the fluctuation amplitude of each dimension of each sub-window in each type of sliding window;

[0021] The average of the fluctuation amplitudes of the three dimensions of the same sub-window is taken as the comprehensive fluctuation amplitude of that sub-window; one comprehensive fluctuation amplitude corresponds to one sub-window.

[0022] Each sub-window is judged based on the comprehensive fluctuation amplitude obtained from each sub-window, thereby identifying abnormal windows;

[0023] The data within the abnormal window is smoothed and corrected to obtain the corrected data of the sub-window. The corrected data of the sub-window and the data of other sub-windows that do not need to be corrected constitute the corrected positioning data.

[0024] Optionally, the structured BeiDou positioning raw dataset includes pseudorange, carrier phase, Doppler shift, satellite azimuth, satellite ID information, and timestamp information;

[0025] The step of performing time synchronization correction on the corrected location data to obtain a time-synchronized location dataset includes:

[0026] The length of the satellite signal propagation path is obtained from the original BeiDou positioning dataset.

[0027] The satellite signal propagation path length is corrected based on the vehicle's altitude information to obtain the corrected satellite signal propagation path length.

[0028] The signal propagation time delay information is calculated based on the electromagnetic wave propagation speed and the corrected satellite signal propagation path length.

[0029] The original timestamp information is corrected based on the signal propagation time delay information to obtain the corrected timestamp information;

[0030] The corrected timestamp information is correlated with other data in the original BeiDou positioning dataset to obtain the time-synchronized positioning dataset.

[0031] Optionally, the step of matching the vehicle's location with its three-dimensional coordinates and an electronic map to obtain the vehicle's location information includes:

[0032] Transform the vehicle's three-dimensional coordinates from the geocentric coordinate system to the Gauss-Kruger plane coordinate system.

[0033] Extract the index information of the environmental feature mapping library from the electronic map;

[0034] Based on the Gauss-Kruger plane coordinates of the vehicle's three-dimensional coordinates and the index information of the environmental feature mapping library in the electronic map, the spatial point-in-polygon algorithm is used to determine the boundary range of the vehicle's coordinates.

[0035] The vehicle's three-dimensional coordinates are corrected based on the boundary range to obtain the corrected vehicle three-dimensional coordinates.

[0036] Optionally, the step of matching the vehicle's location with its three-dimensional coordinates and an electronic map to obtain the vehicle's location information includes:

[0037] Based on the vehicle's three-dimensional coordinates and the electronic map, obtain the vehicle's planar coordinate sequence and spatial data of surrounding candidate roads;

[0038] The degree of alignment between each candidate road and the trajectory segment is obtained based on the vehicle's planar coordinate sequence and the spatial data of surrounding candidate roads.

[0039] The integral of the positional deviation between each candidate road and the trajectory segment is calculated based on the vehicle's planar coordinate sequence and the spatial data of the surrounding candidate roads.

[0040] The specific road identifier of the vehicle is obtained by integrating the degree of alignment between each candidate road and the trajectory segment, as well as the positional deviation between each candidate road and the trajectory segment.

[0041] Optionally, obtaining the vehicle's trajectory information based on the vehicle's location information includes:

[0042] An aligned planar coordinate time series is generated based on the specific road signs where the vehicle is located and the time-synchronized positioning dataset.

[0043] The trajectory curvature is calculated based on the aligned planar coordinate time series, thereby obtaining the curvature value and timestamp corresponding to each positioning point;

[0044] The vehicle's trajectory information is generated based on the curvature value and timestamp corresponding to each positioning point.

[0045] Optionally, the step of determining vehicle violations and identifying dangerous driving behaviors based on vehicle trajectory information includes:

[0046] Obtain real-time environmental data;

[0047] An environmental impact coefficient is generated based on real-time environmental data.

[0048] Obtain the violation rules;

[0049] Information on violations is generated based on violation rules, environmental impact coefficients, and vehicle trajectory information.

[0050] Optionally, the step of determining vehicle violations and identifying dangerous driving behaviors based on vehicle trajectory information further includes:

[0051] The trajectory spatial distribution entropy value of each trajectory window is obtained based on the trajectory information of the vehicle;

[0052] Calculate the rate of change of entropy values ​​for the spatial distribution entropy of the trajectory between every two adjacent trajectory windows;

[0053] Whether dangerous driving behavior exists is determined based on the rate of change of each entropy value.

[0054] Optionally, the trajectory spatial distribution entropy value is calculated using the following formula:

[0055] ;

[0056] in, The entropy value is the spatial distribution of the trajectory. This represents the total number of outermost grid cells. This represents the total number of inner layer grids; The number of trajectory points in the j-th inner grid within the i-th outer grid; This represents the total number of points within the window. Dynamic weights for the inner mesh; This is the window-level adjustment coefficient; This is the minimum value correction coefficient; This is the attenuation factor of the outer grid space.

[0057] This application also provides a vehicle monitoring system based on BeiDou positioning, the vehicle monitoring system based on BeiDou positioning comprising:

[0058] A BeiDou positioning raw data set acquisition module is used to acquire the structured BeiDou positioning raw data set transmitted by BeiDou satellites;

[0059] A quality stratification module is used to perform quality stratification on the acquired BeiDou positioning raw dataset to obtain a positioning dataset with quality level labels, including excellent labels, normal labels and poor labels.

[0060] The correction module is used to initiate multi-window feature comparison and correction on the data corresponding to ordinary labels and inferior labels, so as to obtain the corrected positioning data.

[0061] A time synchronization module is used to perform time synchronization correction on the corrected positioning data, thereby obtaining a time-synchronized positioning dataset.

[0062] A vehicle 3D coordinate acquisition module is used to acquire the vehicle's 3D coordinates based on a time-synchronized positioning dataset.

[0063] An electronic map acquisition module, wherein the electronic map acquisition module is used to acquire electronic maps;

[0064] The vehicle location information acquisition module is used to match the vehicle's location with the vehicle's three-dimensional coordinates and an electronic map, thereby acquiring the vehicle's location information.

[0065] A trajectory information acquisition module, which is used to acquire the trajectory information of a vehicle based on the vehicle's location information;

[0066] The danger assessment module is used to determine the vehicle's violations and identify dangerous driving behaviors based on the vehicle's trajectory information.

[0067] This application has the following advantages:

[0068] (1) Existing technologies for quality screening of location data often rely on a single threshold, which is insufficient to cope with data fluctuations in complex scenarios. This method divides data into excellent, average, and poor quality categories through quality stratification. For average and poor quality data, multi-window (different duration) feature comparison is initiated. Abnormal windows are identified and smoothed by calculating the comprehensive fluctuation amplitude of each window. For example, when the fluctuation amplitude of the 1-second window exceeds 20% of the average of the 3-second and 5-second windows, the average of adjacent three points is used for correction, effectively filtering noise interference and solving the problem of insufficient data correction in existing technologies, thus ensuring the reliability of location data.

[0069] Existing time synchronization correction techniques typically ignore the impact of altitude on signal propagation paths, resulting in significant timestamp discrepancies. This method calculates the propagation path length using satellite azimuth angles, corrects the distance by incorporating vehicle altitude differences (ΔD=h×cosθ), and calculates the time delay based on electromagnetic wave velocity (Δt=D / c), ultimately achieving millimeter-level timestamp compensation (corrected timestamp = original timestamp - Δt_ms + satellite clock bias). This process fully considers altitude factors, overcoming the shortcomings of existing techniques where time synchronization is disconnected from the actual propagation scenario, and ensuring consistency of multi-source data along the timeline.

[0070] Existing technologies often directly compare the original coordinates with a map during location matching, which is susceptible to environmental influences such as building occlusion. This method first converts the 3D coordinates to Gaussian-Kruger plane coordinates, then uses a point-in-polygon algorithm to determine the boundaries of the vehicle's location (e.g., urban canyons, tunnels), and finally combines this with an environmental feature mapping library for targeted deviation correction (e.g., adding lateral correction based on building occlusion angles in urban canyons). Simultaneously, it uses a comprehensive scoring system combining direction fit (60% weight) and location deviation integral (40% weight) to match roads, solving the matching error problem caused by positioning drift in complex environments and improving location matching accuracy in different scenarios.

[0071] Existing technologies use fixed-frequency trajectory sampling, which can easily lead to the loss of key information or data redundancy, and the behavior recognition rules are simplistic. This method dynamically adjusts the sampling frequency based on trajectory curvature (2 seconds for straight sections, 0.5 seconds for turning sections), reducing data volume while preserving key features. Violation judgment incorporates real-time environmental factors (such as a speed limit threshold multiplied by 0.9 in rain or snow), and dangerous driving recognition uses a self-developed entropy formula, fusing multi-dimensional features such as curvature and speed gradient. This design solves the problem of weak correlation between behavior recognition and actual driving scenarios in existing technologies, making the judgment of violations and dangerous behaviors more accurate.

[0072] This method achieves seamless data integration across all stages through a chain design of "quality stratification - time synchronization - location matching - trajectory analysis - behavior recognition": quality correction data provides the foundation for time synchronization, high-precision timestamps ensure the spatiotemporal consistency of location matching, and accurate trajectories provide reliable input for behavior recognition. This collaborative mechanism overcomes the drawbacks of isolated operation of modules in existing technologies, enabling the system to maintain stable performance even in complex scenarios such as urban canyons and tunnels, and possesses strong practical application value. Attached Figure Description

[0073] Figure 1 This is a flowchart illustrating a vehicle monitoring method based on BeiDou positioning according to an embodiment of this application. Detailed Implementation

[0074] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions in the embodiments of this application will be described in more detail below with reference to the accompanying drawings. In the drawings, the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The described embodiments are some, but not all, embodiments of this application. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application. The embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0075] like Figure 1 The vehicle monitoring method based on BeiDou positioning shown includes:

[0076] Step 1: Obtain the structured raw dataset of BeiDou positioning transmitted by BeiDou satellites;

[0077] Step 2: Perform quality stratification on the acquired BeiDou positioning raw dataset to obtain a positioning dataset with quality level labels, including excellent labels, normal labels and poor labels;

[0078] Step 3: Perform multi-window feature comparison and correction on the data corresponding to ordinary labels and inferior labels to obtain the corrected location data;

[0079] Step 4: Perform time synchronization correction on the corrected location data to obtain the time-synchronized location dataset;

[0080] Step 5: Obtain the vehicle's 3D coordinates based on the time-synchronized positioning dataset;

[0081] Step 6: Obtain the electronic map;

[0082] Step 7: Match the vehicle's location with the vehicle's 3D coordinates and the electronic map to obtain the vehicle's location information;

[0083] Step 8: Obtain the vehicle's trajectory information based on the vehicle's location information;

[0084] Step 9: Determine vehicle violations and identify dangerous driving behaviors based on vehicle trajectory information.

[0085] In this embodiment, the monitored vehicle may need to be equipped with a receiving module. For example, an industrial-grade Beidou dual-mode (compatible with BDS B1I / B2I frequency bands) receiving module can be used to meet the requirements of high-dynamic vehicle scenarios.

[0086] In this embodiment, the signal capture principle is as follows:

[0087] Visible satellites are identified by searching for the carrier frequency (affected by Doppler shift) and pseudocode phase (the BeiDou B1I signal uses C / A code with a code length of 1ms and a code rate of 2.046Mcps).

[0088] FFT (Fast Fourier Transform) is used to accelerate acquisition. The input signal is segmented (every 1ms segment) and correlated with the locally generated pseudocode. When the correlation peak exceeds the threshold (such as 6 times the noise mean), the acquisition is considered successful.

[0089] The carrier frequency deviation is eliminated by a PLL (phase-locked loop) to output the carrier phase (accuracy up to 0.1 cycles); when the vehicle is moving at high speed (such as 120 km / h), an adaptive bandwidth PLL (bandwidth 5-20 Hz) is activated to dynamically compensate for Doppler frequency shift changes.

[0090] The phase of the local pseudocode and the received pseudocode are aligned by a DLL (Delay Locked Loop) to output the pseudorange measurement value (error ≤ 1m); a narrow correlation technique (correlation spacing 0.1 chip) is used to suppress pseudorange deviation caused by multipath signals.

[0091] In this embodiment, the following core parameters can be parsed and extracted according to a preset protocol:

[0092] Satellite identifier (PRN code, such as C01-C05 for BeiDou GEO satellites);

[0093] Pseudorange (unit: m, raw distance including atmospheric delay, clock error, and other errors);

[0094] Carrier phase (unit: cycles, integer cycles + fractional cycles, accuracy up to 0.01 cycles);

[0095] Doppler shift (unit: Hz, reflecting the relative speed of the satellite and the vehicle);

[0096] Timestamp (receiver local time, format UTC + milliseconds);

[0097] Other information.

[0098] In this embodiment, the acquired raw BeiDou positioning dataset is stratified by quality to obtain a positioning dataset with quality level labels. These quality level labels include excellent, average, and poor labels.

[0099] The raw BeiDou positioning dataset is preprocessed, specifically as follows:

[0100] Traverse the original BeiDou positioning dataset and extract the signal strength (unit: dBm) and signal-to-noise ratio (unit: dB) fields row by row;

[0101] Filter out null or outlier values ​​(such as obviously erroneous data with signal strength > -50dBm or < -150dBm).

[0102] A temporary index is created on the cleaned data and associated with other parameters (satellite ID, timestamp, etc.) in the original data.

[0103] Define the quality stratification judgment rules, for example, as follows:

[0104] Excellent: Signal strength ≥ -85dBm and signal-to-noise ratio ≥ 45dB;

[0105] Normal: -100dBm ≤ signal strength < -85dBm and 30dB ≤ signal-to-noise ratio < 45dB;

[0106] Poor quality: Signal strength < -100dBm or signal-to-noise ratio < 30dB (meeting either condition is sufficient for judgment).

[0107] In some cases, special circumstances may arise. For example, when the signal strength meets the normal requirements but the signal-to-noise ratio is poor, the worst setting will be used.

[0108] The quality level labels after stratification are associated with other original fields in the structured BeiDou positioning raw dataset, and sorted in ascending order by timestamp to generate a positioning dataset with quality level labels.

[0109] In this embodiment, the step of performing multi-window feature comparison and correction on the data corresponding to ordinary tags and inferior tags to obtain corrected location data includes:

[0110] From the location dataset with quality grade labels, select the data corresponding to ordinary labels and poor-quality labels as the data to be corrected;

[0111] Add a timestamp index to each piece of data to be corrected to ensure that the data is arranged in chronological order;

[0112] Create multiple independent sliding windows and initialize and configure the parameters of each type of sliding window. Each type of sliding window has a different window duration than the other types of sliding windows, and the window duration of each child window within each type of sliding window is the same. For example, there can be three types of sliding windows: the first type of sliding window is set to 1 second, the second type of sliding window is set to 3 seconds, and the third type of sliding window is set to 5 seconds (it can be understood that the actual settings can be adjusted according to the needs).

[0113] The coordinate values ​​of longitude, latitude, and altitude are extracted from the data to be corrected using each type of sliding window.

[0114] Calculate the fluctuation amplitude for each dimension of each sub-window within each type of sliding window; for example, latitude fluctuation amplitude = maximum latitude value within the window - minimum latitude value).

[0115] The average of the fluctuation amplitudes of the three dimensions of the same sub-window is taken as the comprehensive fluctuation amplitude of the sub-window; one comprehensive fluctuation amplitude corresponds to one sub-window; for example, for the same sub-window, the latitude fluctuation amplitude is 2, the longitude fluctuation amplitude is 3, and the altitude fluctuation amplitude is 4, then the comprehensive fluctuation amplitude = (2+3+4) / 3, which is 3.

[0116] Each sub-window is judged based on the comprehensive fluctuation amplitude obtained from each sub-window to identify abnormal windows. Specifically, for three windows within the same time interval, the average comprehensive fluctuation amplitude of two of the windows is calculated (e.g., the average of the 1-second window and the 3-second window is calculated as a reference benchmark for the 5-second window). When the comprehensive fluctuation amplitude of a certain window exceeds 20% of the average of the other two windows, the window is marked as an "abnormal window".

[0117] For example, suppose the current time is t = 10.0 seconds (with this time as the endpoint, three windows are captured backwards), and the specific values ​​are shown below:

[0118] 1. Time range of the three types of windows (ending at t=10.0 seconds):

[0119] 1-second window: Coverage time 9.0 seconds to 10.0 seconds (total 1 second);

[0120] 3-second window: Coverage time 7.0 seconds to 10.0 seconds (total 3 seconds);

[0121] 5-second window: Coverage time 5.0 seconds to 10.0 seconds (total 5 seconds);

[0122] The three windows all end at 10.0 seconds (the same endpoint), but cover different historical durations. Therefore, they can be regarded as three windows within the same time interval (the interval ending at 10.0 seconds).

[0123] 2. Calculation of the overall fluctuation range for each window

[0124] Assume the location data (latitude) fluctuates as follows within the three windows (latitude is used as an example only; the actual fluctuation is a combination of longitude, latitude, and altitude):

[0125] 1-second window (9.0~10.0 seconds):

[0126] Latitude values: 30.0000° → 30.0002° → 30.0001°;

[0127] Fluctuation range (maximum value - minimum value): 30.0002° - 30.0000° = 0.0002°;

[0128] The overall fluctuation range A = 0.0002°;

[0129] 3-second window (7.0~10.0 seconds):

[0130] Latitude values: 30.0000°→30.0003°→30.0001°→30.0002°→30.0000°;

[0131] The fluctuation range is 30.0003° - 30.0000 = 0.0003°;

[0132] The overall fluctuation range B = 0.0003°.

[0133] 5-second window (5.0~10.0 seconds):

[0134] Latitude values: 30.0000° → 30.0005° → 30.0002° → 30.0006° → 30.0001° → 30.0003°;

[0135] Fluctuation range: 30.0006° - 30.0000° = 0.0006°;

[0136] The overall fluctuation range is C = 0.0006°.

[0137] 3. Abnormal window detection (taking a 5-second window as an example)

[0138] Calculate the average combined fluctuation amplitude for the 1-second and 3-second windows:

[0139] Reference datum = (A + B) ÷ 2 = (0.0002° + 0.0003°) ÷ 2 = 0.00025°;

[0140] Determine if the 5-second window is abnormal:

[0141] The difference between the 5-second window fluctuation amplitude C (0.0006°) and the reference baseline (0.00025°) is:

[0142] 0.0006° - 0.00025° = 0.00035°;

[0143] The deviation ratio = (0.00035° ÷ 0.00025°) × 100% = 140%, which far exceeds the preset threshold of 20%.

[0144] Conclusion: The 5-second window (5.0~10.0 seconds) was identified as an "abnormal window".

[0145] The data within the anomaly window is smoothed and corrected to obtain the corrected data for the sub-window. This corrected data, along with the data from other sub-windows that do not require correction, constitutes the corrected positioning data. For example, for a single anomaly point, the coordinates of its preceding and following valid points are extracted (if there are no valid points on one side, the two nearest points on the same side are used). The arithmetic mean of the three points (one before and one after the anomaly point plus the anomaly point itself) in longitude, latitude, and altitude is calculated as the corrected coordinate value. For consecutive anomalies (no more than three), segmented correction is used: the first anomaly point is corrected using the mean of the preceding and following two points, and the second anomaly point is corrected using the mean of the preceding and following two points.

[0146] Using this method, by anchoring to the same endpoint time (e.g., 10.0 seconds), although the three windows cover different historical durations, the average fluctuation of the "short-term + medium-term" windows can be used as a benchmark to determine whether the fluctuation of the "long-term window" is abnormal. Similarly, the average of the 1-second and 5-second windows can be used to judge the 3-second window, or the average of the 3-second and 5-second windows can be used to judge the 1-second window, thus achieving multi-dimensional anomaly verification.

[0147] It can be understood that when there are no multiple types of windows at a given time point, the data can be assumed to be normal. For example, if there is no 5-second window within the time point of 0-3 seconds, the first 4 seconds will be assumed to be normal data.

[0148] In this embodiment, the window judgment adopts the design of "the same endpoint" to ensure that windows of different durations always focus on "the fluctuations and recent trends at the current time point", thus ensuring the real-time, relevance and accuracy of anomaly judgment.

[0149] In this embodiment, the structured BeiDou positioning raw dataset includes pseudorange, carrier phase, Doppler shift, satellite azimuth, satellite ID information, and timestamp information;

[0150] In this embodiment, the step of performing time synchronization correction on the corrected location data to obtain a time-synchronized location dataset includes:

[0151] The length of the satellite signal propagation path is obtained from the original BeiDou positioning dataset.

[0152] In this embodiment, the satellite signal propagation path length (slant range) is calculated using the law of cosines:

[0153] ;

[0154] Where θ is the satellite elevation angle; R is the average radius of the Earth; S is the distance from the BeiDou satellite to the Earth's center (a known constant); D0 is the length of the satellite signal propagation path; and 90°-θ is the angle between the line connecting the satellite and the Earth's center and the Earth's radius.

[0155] The satellite signal propagation path length is corrected based on the vehicle's altitude information to obtain the corrected satellite signal propagation path length. Specifically, the propagation distance correction ΔD caused by the vehicle's altitude is calculated, where ΔD = h × cosθ (h is the vehicle's altitude; θ is the satellite elevation angle). The corrected satellite signal propagation path length D = D0 - ΔD (because the actual propagation distance is shorter when the vehicle is above the ellipsoid than at the ellipsoid).

[0156] The signal propagation time delay is calculated based on the electromagnetic wave propagation speed and the corrected satellite signal propagation path length; specifically, based on the electromagnetic wave propagation speed c = 3 × 10⁻⁶. 8 Given the speed of m / s, the propagation time delay Δt is calculated using the following formula:

[0157] ;

[0158] Where Δt is the propagation time delay (in seconds); D is the corrected satellite signal propagation path length; and c is the electromagnetic wave propagation speed.

[0159] The original timestamp information is corrected based on the signal propagation time delay information to obtain the corrected timestamp information;

[0160] In this embodiment, the corrected timestamp information is obtained using the following formula:

[0161] Corrected timestamp = Original received timestamp − Δt + Satellite clock bias.

[0162] The corrected timestamp information is correlated with other data in the original BeiDou positioning dataset to obtain the time-synchronized positioning dataset.

[0163] In this embodiment, the vehicle's three-dimensional coordinates can be obtained from the time-synchronized positioning dataset using the following method:

[0164] Extract pseudorange, corrected timestamp, and other information of the satellites transmitting the original BeiDou positioning dataset; parse the ephemeris to obtain the satellite's three-dimensional coordinates (X, Y, Z) and clock difference; retrieve the vehicle's most recent valid historical three-dimensional coordinates (x0, y0, z0).

[0165] Using the pseudorange equation, the historical position (x0, y0, z0) is taken as the initial value. Assuming that the vehicle position changes little in a short period of time, the equation is solved by iterative method, and the coordinate values ​​are gradually corrected until the deviation between the calculated pseudorange and the measured pseudorange is less than a threshold (such as 1 meter), thereby obtaining the vehicle's three-dimensional coordinates.

[0166] In this embodiment, the step of matching the vehicle's location based on its three-dimensional coordinates and an electronic map to obtain the vehicle's location information includes:

[0167] The vehicle's three-dimensional coordinates are converted from the geocentric-fixed coordinate system to Gauss-Kruger plane coordinates. In this embodiment, the initial three-dimensional coordinates are converted from the geocentric-fixed coordinate system to Gauss-Kruger plane coordinates (X, Y) to facilitate matching with the plane coordinate system of the environmental feature mapping library. The index information of the environmental feature mapping library (such as the latitude and longitude range of the region boundary and the feature label encoding) is extracted.

[0168] Extract the index information of the environmental feature mapping library from the electronic map;

[0169] In this embodiment, an environmental feature mapping library is first obtained through an electronic map. The specific method is as follows:

[0170] Retrieve structured data from electronic map APIs (such as Baidu Maps Open Platform and Google Maps API). For example, structured data includes:

[0171] Road network data: road class (expressway, arterial road, secondary arterial road), number of lanes, direction of traffic, speed limit.

[0172] Point of Interest (POI) data: building outline (polygon coordinates), building type (residential, commercial, industrial), landmark name and coordinates (e.g., tunnel entrance, bridge).

[0173] Topographic elevation data: Elevation values ​​of contour lines or discrete points provided by some maps (accuracy is usually 1-10 meters).

[0174] The WGS84 latitude and longitude coordinates returned by the electronic map are uniformly converted into Gauss-Kruger plane coordinates to ensure consistency in spatial location calculations.

[0175] The POI data is deduplicated (removing duplicate annotations of the same building) and stored in categories such as "region-feature type" (e.g., "city center-commercial buildings" and "suburbs-industrial buildings").

[0176] Each area is classified. For example, based on the density calculation of building POIs (number of buildings per unit area), if there are ≥50 building POIs within 1 square kilometer and the road width is ≤20 meters, it is identified as a potential urban canyon area.

[0177] Alternatively, you can directly call the POIs marked "tunnel" or "bridge" on the map, extract their entrance / exit coordinate range, and use them as the boundary of the special environmental area.

[0178] By storing the above data, an environmental feature mapping library can be formed.

[0179] Based on the Gauss-Kruger plane coordinates of the vehicle's three-dimensional coordinates and the index information of the environmental feature mapping library in the electronic map, the spatial point-in-polygon algorithm is used to determine the boundary range of the vehicle's coordinates.

[0180] In this embodiment, the spatial point-in-polygon algorithm determines whether a point is inside a polygon by judging the positional relationship between the vehicle coordinates (point) and the region boundary (polygon). A ray is emitted from the vehicle coordinates in any direction (e.g., horizontally to the right), and the number of intersections between the ray and the polygon boundary is counted. If the number of intersections is odd, the point is inside the polygon; if the number is even, the point is outside the polygon.

[0181] Extract the polygon boundary coordinates of the target region from the environmental feature mapping library (such as the polygon vertex coordinate sequence of an urban canyon area), and arrange them in clockwise or counterclockwise order.

[0182] Obtain the vehicle's planar coordinates (X, Y) as the spatial point to be judged.

[0183] Initialize the intersection count to 0. Traverse each edge of the polygon (the line segment formed by two adjacent vertices). For each edge, determine whether the ray (horizontally to the right from the vehicle coordinates) intersects the edge. If they intersect, increment the intersection count by 1. After traversal, if the intersection count is odd, determine that the vehicle is in the region; if it is even, determine that it is outside the region.

[0184] For example, suppose we have the following known conditions:

[0185] The coordinates of the polygonal boundary vertices of the urban canyon area (arranged clockwise) are: A (100, 200), B (150, 200), C (150, 300), D (100, 300) (unit: meters, Gauss-Kruger plane coordinates).

[0186] The vehicle's planar coordinates are P(120, 250).

[0187] Determine the four sides of the polygon: AB(100,200)-(150,200), BC(150,200)-(150,300), CD(150,300)-(100,300), DA(100,300)-(100,200).

[0188] From the vehicle coordinates P(120, 250), emit a ray horizontally to the right (y=250, x≥120), and determine the intersection points of the ray with the edges one by one:

[0189] Side AB: y=200, ray y=250, they do not intersect.

[0190] Side BC: x=150, y ranges from 200 to 300, the ray intersects side BC at (150, 250), the intersection count is 1.

[0191] Edge CD: y=300, ray y=250, they do not intersect.

[0192] Edge DA: x=100, y ranges from 300 to 200, ray x≥120, non-intersecting.

[0193] If the intersection count is 1 (an odd number), it is determined that vehicle P is within the city canyon area.

[0194] The vehicle's three-dimensional coordinates are corrected based on the boundary range to obtain the corrected vehicle three-dimensional coordinates.

[0195] In this embodiment, the correction methods differ for different areas. For example, in urban canyon areas, the deviation may be caused by building obstruction angles, so a deviation coefficient is set for the building obstruction angles for correction. In tunnel areas, the deviation may be caused by electromagnetic interference intensity levels, so compensation is set according to the electromagnetic interference intensity levels.

[0196] For example:

[0197] Urban canyon area: Obtain the building occlusion angle α (the angle between the line connecting the vehicle and the building apex and the horizontal line) and the lateral deviation coefficient k1 (0.02-0.08 meters / degree, increasing with α).

[0198] Lateral deviation ΔX = α × k1 (α is the occlusion angle, k1 is the lateral deviation coefficient), and the correction formula is X correction = X initial + ΔX × sign (building orientation) (the sign value is ±1 depending on whether the building is on the left or right side of the vehicle).

[0199] Tunnel area: Obtain the electromagnetic interference intensity level (level 1-5) and match the longitudinal deviation compensation value ΔL (level 1 corresponds to +0.5 meters, and each level increases by 0.3 meters).

[0200] Longitudinal deviation ΔY = ΔL (electromagnetic interference compensation value), correction formula is Y correction = Y initial + ΔY; Altitude deviation ΔH = 0.1 × h_tunnel (10% of tunnel height h_tunnel, due to air pressure error)

[0201] Comprehensive correction: The deviations in each dimension are superimposed to obtain the corrected three-dimensional coordinates.

[0202] Understandably, various correction methods can be set as needed.

[0203] In this embodiment, the step of matching the vehicle's location based on its three-dimensional coordinates and an electronic map to obtain the vehicle's location information includes:

[0204] Based on the vehicle's three-dimensional coordinates and the electronic map, obtain the vehicle's planar coordinate sequence and spatial data of surrounding candidate roads; specifically, convert the vehicle's three-dimensional coordinates into planar coordinates (X, Y), and use data from the electronic map... Figure 1 Establish a consistent coordinate system (such as the Gauss-Kruger coordinate system); extract candidate road data within a 500-meter radius of the vehicle's current location from the dynamic electronic map, including the centerline coordinate sequence, road width, lane boundary coordinates, etc. of each road.

[0205] The degree of alignment between each candidate road and the trajectory segment is obtained based on the vehicle's planar coordinate sequence and the spatial data of surrounding candidate roads. Specifically, five consecutive vehicle positioning points (the number can be set as needed) are extracted to form a trajectory segment, and the time interval of the trajectory segment is calculated (to ensure uniform sampling, such as one point every 0.5 seconds). Based on the straight-line distance between the vehicle position and the road, the candidate roads closest to the trajectory segment are selected (distance ≤ 50 meters, excluding roads that are obviously far away).

[0206] Linear fitting is performed on 5 consecutive vehicle positioning points to obtain the tangent direction vector of the trajectory (such as the direction vectors of points 1-2, 2-3, 3-4, and 4-5, and the average value is taken as the overall trajectory direction).

[0207] Extract the coordinate sequence of the centerline of each candidate road, take a 200-meter road segment close to the trajectory position, and perform linear fitting on it to obtain the tangent direction vector of the road.

[0208] The alignment degree is calculated by the angle θ between the trajectory tangent direction vector and the road tangent direction vector. The formula is: Alignment degree = 1 - θ / 90° (θ is in degrees, ranging from 0° to 90°, and the alignment degree ranges from 0 to 1; the larger the value, the higher the alignment degree). If the angle is greater than 90°, the supplementary angle is used for calculation (because there may be alignment differences between the two directions of the road).

[0209] The integral of the positional deviation between each candidate road and the trajectory segment is calculated based on the vehicle's planar coordinate sequence and the spatial data of the surrounding candidate roads.

[0210] Specifically, for each location point of the trajectory segment, calculate its vertical distance (d_i, i=1-5) to the center line of the candidate road. If the point is on the left side of the road, the distance is positive, and if it is on the right side, the distance is negative (for subsequent direction determination).

[0211] Sum the absolute values ​​of the distances between the 5 points, i.e., the deviation integral = Σ|d_i| (i = 1 to 5), and then divide by the length of the trajectory segment (the total length of the broken line composed of the 5 points) to obtain the deviation integral per unit length (the smaller the value, the closer the position is to the road).

[0212] The specific road identifier of the vehicle is obtained based on the degree of alignment between each candidate road and the trajectory segment and the positional deviation integral between each candidate road and the trajectory segment. Specifically, a weight can be set first, for example, the alignment degree weight is 60% and the positional deviation integral weight is 40% (the deviation integral needs to be normalized and converted into a "deviation score" between 0 and 1, that is, deviation score = 1 - deviation integral / maximum deviation integral, and the maximum deviation integral is set according to the maximum value among the candidate roads).

[0213] Next, calculate the overall score using the following formula: Overall score = Direction matching degree × 60% + Deviation score × 40%.

[0214] The 10 candidate roads are ranked by their overall scores, and the road with the highest score is selected as the matching result.

[0215] In this embodiment, obtaining the vehicle's trajectory information based on the vehicle's location information includes:

[0216] Based on the specific road signs where the vehicle is located and the time-synchronized positioning dataset, an aligned planar coordinate time series is generated. Specifically, the real-time positioning coordinate sequence corresponding to the real-time vehicle location information is converted into planar coordinates (X, Y) and spatially aligned with the coordinates of the road centerline of the driving path. The timestamp information in the coordinate sequence is extracted, and the time interval between adjacent positioning points is calculated (to ensure data continuity and remove abnormal breakpoints with time intervals > 1 second).

[0217] The trajectory curvature is calculated based on the aligned planar coordinate time series, thereby obtaining the curvature value and timestamp corresponding to each positioning point;

[0218] In this embodiment, trajectory curvature calculation is performed based on the aligned planar coordinate time series to obtain the curvature value and timestamp corresponding to each positioning point, including:

[0219] The calculation window is formed by taking one point before and one point after the current location point as the center, and the window slides with the time series.

[0220] The curvature of the arc is calculated by fitting the coordinates of three points to the arc (curvature = 1 / turning radius). The formula is:

[0221] Let the coordinates of the three points be A(x1,y1), B(x2,y2), and C(x3,y3). First, calculate the slopes of line segments AB and BC (the slope of AB is k1=(y2-y1) / (x2-x1), and the slope of line segment BC is k2=(y3-y2) / (x3-x2).

[0222] The turning angle Δθ is then calculated using the included angle formula (Δθ=|arctan(k2)-arctan(k1)|×(180 / π)). Combined with the line segment length L (the average length of AB and BC), the curvature k=Δθ / L is obtained. The average curvature values ​​of three consecutive windows are taken to reduce single-point calculation errors.

[0223] The vehicle's trajectory information is generated based on the curvature value and timestamp corresponding to each positioning point.

[0224] For example, when the curvature values ​​of three consecutive positioning points are all <0.01° / meter, the trajectory is determined to be a straight line segment.

[0225] When the curvature value of any positioning point is ≥0.01° / meter, or when the curvature of at least 2 out of 3 consecutive points is ≥0.01° / meter, the trajectory segment is determined to be a turning / lane change segment.

[0226] At the intersection of the straight segment and the turning segment, extend two points forward and two points backward to ensure the continuous division of the trajectory segment (if the curvature of the last two points of the straight segment is close to the threshold, they are included in the calculation of the turning segment).

[0227] In this embodiment, different sampling methods can be used when sampling locations depending on the road segment. For example, for a trajectory marked as a "straight line segment", sampling is performed at 2-second intervals, that is, starting from the starting point of the segment, one positioning point is extracted every 2 seconds (the specific situation can be set as needed).

[0228] For trajectories marked as "turning / lane change sections", switch to 0.5-second interval sampling, that is, extract one positioning point every 0.5 seconds (when sampling at the original 1Hz, each point is retained; when sampling at higher frequencies, it is extracted proportionally).

[0229] When transitioning from a straight section to a turning section, the sampling frequency is set to the straight section frequency for the first second, and then switched to the turning section frequency in advance for the next second to avoid losing critical nodes.

[0230] For each sampling point, key features are added, including the curvature value of the point, the corresponding road ID, the lane position, the distance (ΔS) to the previous sampling point, and the time difference (Δt).

[0231] In the sampling points of the turning / lane change section, the steering direction (left / right, determined by the positive or negative curvature) and instantaneous speed (calculated by ΔS / Δt) are additionally recorded.

[0232] If the coordinate deviation of consecutive sampling points is less than 0.5 meters (e.g., when the vehicle is briefly stopped), only the first and last points are retained to avoid redundant data.

[0233] In this embodiment, the step of determining vehicle violations and identifying dangerous driving behaviors based on vehicle trajectory information includes:

[0234] Obtain real-time environmental data;

[0235] An environmental impact coefficient is generated based on real-time environmental data. For example, real-time environmental data can include weather factors and time-of-day factors. In one embodiment, the following settings can be made:

[0236] The coefficient for sunny weather is 1.0; the speed limit threshold coefficient is adjusted to 0.9 during rainy or snowy weather.

[0237] The coefficient for normal hours is 1.0; during school hours (such as 7:30-8:30 and 16:30-17:30), the threshold coefficient for determining low-speed lane occupation is adjusted to 0.8.

[0238] For situations where multiple environmental factors coexist, the product of the coefficients is taken (e.g., rainy / snowy weather + school hours, speed limit threshold coefficient = 0.9 × 0.8 = 0.72).

[0239] Obtain the violation rules;

[0240] In this embodiment, the violation rules can be set as needed. Violation rules in traffic law can be used, or other violation rules can be set by oneself, such as speeding violation judgment, low-speed lane occupation violation judgment, and lane violation judgment.

[0241] Among them, when the instantaneous speed of a vehicle is greater than the road speed limit multiplied by the speed limit coefficient, it is judged as speeding (e.g., if the speed limit is 60km / h, the rain and snow weather coefficient is 0.9, and the actual speed is greater than 54km / h, it is considered speeding).

[0242] On highways or main roads, in the fast lane (leftmost lane 1-2), if the speed is less than the minimum speed limit multiplied by the low-speed lane occupation coefficient (e.g., minimum speed limit 60km / h, school time coefficient 0.8, speed < 48km / h for more than 10 seconds), it is considered low-speed lane occupation.

[0243] Based on the lane traffic direction in the dynamic electronic map, if the trajectory node shows that a vehicle is in a prohibited lane (such as going against traffic or going straight in a right-turn lane), it is determined to be a lane violation.

[0244] Information on violations is generated based on violation rules, environmental impact coefficients, and vehicle trajectory information.

[0245] Specifically, when the information displayed by a vehicle's trajectory exceeds the violation rules, violation information is generated.

[0246] Understandably, some special judgment rules can also be set, such as short-term anomaly filtering and continuous violations.

[0247] For example, if the violation only lasts for one node (e.g., within 0.5 seconds), and the nodes before and after it are normal, it is judged as a measurement error and the record is removed.

[0248] Alternatively, consecutive violations of the same type (such as multiple nodes exceeding the speed limit consecutively) can be merged into a single record, recording the start and end times.

[0249] In this embodiment, the step of determining vehicle violations and identifying dangerous driving behaviors based on vehicle trajectory information further includes:

[0250] The trajectory spatial distribution entropy value of each trajectory window is obtained based on the trajectory information of the vehicle. Specifically, a sliding window with a preset number of seconds (e.g., 10 seconds) can be set, with each window containing 20 points (based on a 0.5-second sampling interval), and the window slides over time (moving 1 second each time).

[0251] Divide the trajectory space within the window into a 1m x 1m grid (you can set it as needed), and count the number of trajectory points in each grid.

[0252] Calculate the rate of change of entropy values ​​for the spatial distribution entropy of the trajectory between every two adjacent trajectory windows;

[0253] In this embodiment, the trajectory spatial distribution entropy value is calculated using the following formula:

[0254] ;

[0255] in, The entropy value is the spatial distribution of the trajectory. This represents the total number of outermost grid cells. This represents the total number of inner layer grids; The number of trajectory points in the j-th inner grid within the i-th outer grid; This represents the total number of points within the window. Dynamic weights for the inner mesh; This is the window-level adjustment coefficient; This is the minimum value correction coefficient; This is the attenuation factor of the outer grid space.

[0256] Using the entropy calculation formula of this application, the grid scale is dynamically adjusted with speed, which can solve the problem of insufficient adaptability of traditional fixed grids in high and low speed scenarios; and the introduction of curvature weight makes the trajectory dispersion of dangerous scenarios such as sharp bends have a more significant impact on the entropy value; finally, combined with spatiotemporal characteristics, it can better reflect the dynamic risks of driving behavior than simple spatial distribution entropy.

[0257] The rate of change of entropy between two adjacent windows is calculated using the formula: "Rate of change = (current window entropy - previous window entropy) / previous window entropy × 100%".

[0258] Whether dangerous driving behavior exists is determined based on the rate of change of each entropy value.

[0259] For example, when the entropy change rate is greater than 50% within 10 seconds, the type can be further determined by combining curvature and speed (such as a sudden increase in curvature + a sudden increase in entropy value indicates a dangerous driving behavior of sudden turn, and a sudden change in lane position + a sudden increase in entropy value indicates a dangerous driving behavior of sudden lane change).

[0260] This application has the following advantages:

[0261] (1) Existing technologies for quality screening of location data often rely on a single threshold, which is difficult to cope with data fluctuations in complex scenarios. This method divides data into excellent, normal, and poor quality labels through quality stratification. For normal and poor quality data, multi-window (different duration) feature comparison is initiated. Abnormal windows are identified and smoothed by calculating the comprehensive fluctuation amplitude of each window. For example, when the fluctuation amplitude of the 1-second window exceeds 20% of the average of the 3-second and 5-second windows, the average of adjacent three points is used for correction, which effectively filters noise interference and solves the problem of insufficient data correction in existing technologies, ensuring the reliability of location data.

[0262] Existing time synchronization correction techniques typically ignore the impact of altitude on signal propagation paths, resulting in significant timestamp discrepancies. This method calculates the propagation path length using satellite azimuth angles, corrects the distance by incorporating vehicle altitude differences (ΔD = h × cosθ), and calculates the time delay based on electromagnetic wave velocity (Δt = D / c), ultimately achieving millimeter-level timestamp compensation (corrected timestamp = original timestamp - Δt_ms + satellite clock bias). This process fully considers altitude factors, overcoming the defect in existing techniques where time synchronization is disconnected from the actual propagation scenario, and ensuring consistency of multi-source data along the timeline.

[0263] Existing technologies often directly compare the original coordinates with a map during location matching, which is susceptible to environmental influences such as building occlusion. This method first converts the 3D coordinates to Gaussian-Kruger plane coordinates, then uses a point-in-polygon algorithm to determine the boundaries of the vehicle's location (e.g., urban canyons, tunnels), and finally combines this with an environmental feature mapping library for targeted deviation correction (e.g., adding lateral correction based on building occlusion angles in urban canyons). Simultaneously, by comprehensively scoring the road matching using both the alignment accuracy (60% weight) and the location deviation integral (40% weight), the method solves the matching error problem caused by positioning drift in complex environments, improving location matching accuracy in different scenarios.

[0264] Existing technologies use fixed-frequency trajectory sampling, which can easily lead to the loss of key information or data redundancy, and the behavior recognition rules are simplistic. This method dynamically adjusts the sampling frequency based on trajectory curvature (2 seconds for straight sections, 0.5 seconds for turning sections), reducing data volume while preserving key features. Violation judgment incorporates real-time environmental factors (such as a speed limit threshold multiplied by 0.9 in rain or snow), and dangerous driving recognition uses a self-developed entropy formula, fusing multi-dimensional features such as curvature and speed gradient. This design solves the problem of weak correlation between behavior recognition and actual driving scenarios in existing technologies, making the judgment of violations and dangerous behaviors more accurate.

[0265] This method achieves seamless data integration across all stages through a chain design of "quality stratification - time synchronization - location matching - trajectory analysis - behavior recognition": quality correction data provides the foundation for time synchronization, high-precision timestamps ensure the spatiotemporal consistency of location matching, and accurate trajectories provide reliable input for behavior recognition. This collaborative mechanism overcomes the drawbacks of isolated operation of modules in existing technologies, enabling the system to maintain stable performance even in complex scenarios such as urban canyons and tunnels, and possesses strong practical application value.

[0266] This application also provides a vehicle monitoring system based on BeiDou positioning. The BeiDou-based vehicle monitoring system includes a BeiDou positioning raw dataset acquisition module, a quality stratification module, a correction module, a time synchronization module, a vehicle three-dimensional coordinate acquisition module, an electronic map acquisition module, a vehicle location information acquisition module, a trajectory information acquisition module, and a hazard judgment module.

[0267] The BeiDou positioning raw dataset acquisition module is used to acquire structured BeiDou positioning raw datasets transmitted by BeiDou satellites.

[0268] The quality stratification module is used to perform quality stratification on the acquired BeiDou positioning raw dataset, thereby obtaining a positioning dataset with quality level labels, which include excellent labels, normal labels and poor labels.

[0269] The correction module is used to initiate multi-window feature comparison and correction for the data corresponding to ordinary labels and inferior labels, thereby obtaining the corrected location data;

[0270] The time synchronization module is used to perform time synchronization correction on the corrected location data, thereby obtaining a time-synchronized location dataset;

[0271] The vehicle 3D coordinate acquisition module is used to acquire the vehicle's 3D coordinates based on the time-synchronized positioning dataset;

[0272] The electronic map acquisition module is used to acquire electronic maps;

[0273] The vehicle location information acquisition module is used to match the vehicle's three-dimensional coordinates with the electronic map to obtain the vehicle's location information.

[0274] The trajectory information acquisition module is used to acquire the vehicle's trajectory information based on the vehicle's location information;

[0275] The hazard assessment module is used to determine vehicle violations and identify dangerous driving behaviors based on the vehicle's trajectory information.

[0276] Although the present invention has been described in detail above with general descriptions and specific embodiments, modifications or improvements can be made to it, which will be obvious to those skilled in the art. Therefore, all such modifications or improvements made without departing from the spirit of the present invention fall within the scope of protection claimed by the present invention.

Claims

1. A vehicle monitoring method based on BeiDou positioning, characterized in that, The vehicle monitoring method based on BeiDou positioning includes: Obtain the structured raw dataset of BeiDou positioning transmitted by BeiDou satellites; The acquired BeiDou positioning raw dataset is stratified by quality to obtain a positioning dataset with quality level labels, which include excellent labels, normal labels and poor labels. Multi-window feature comparison and correction are initiated for the data corresponding to ordinary labels and inferior labels to obtain corrected location data; The corrected location data is time-synchronized and corrected to obtain a time-synchronized location dataset. Obtain the vehicle's 3D coordinates from the time-synchronized positioning dataset; Obtain electronic maps; Vehicle location information is obtained by matching the vehicle's three-dimensional coordinates with an electronic map. Obtain vehicle trajectory information based on vehicle location information; Based on the vehicle's trajectory information, determine the vehicle's violations and identify dangerous driving behaviors; The process of performing multi-window feature comparison and correction on the data corresponding to ordinary and inferior labels to obtain corrected location data includes: From the location dataset with quality grade labels, select the data corresponding to ordinary labels and poor-quality labels as the data to be corrected; Add a timestamp index to each piece of data to be corrected; Create multiple independent sliding windows and initialize and configure the parameters of each type of sliding window. Each type of sliding window has a different window duration than other types of sliding windows, and each child window within each type of sliding window has the same window duration. The coordinate values ​​of longitude, latitude, and altitude are extracted from the data to be corrected using each type of sliding window. Calculate the fluctuation amplitude of each dimension of each sub-window in each type of sliding window; The average of the fluctuation amplitudes of the three dimensions of the same sub-window is taken as the comprehensive fluctuation amplitude of that sub-window; one comprehensive fluctuation amplitude corresponds to one sub-window. Each sub-window is judged based on the comprehensive fluctuation amplitude obtained from each sub-window, thereby identifying abnormal windows; The data within the abnormal window is smoothed and corrected to obtain the corrected data of the sub-window. The corrected data of the sub-window and the data of other sub-windows that do not need to be corrected constitute the corrected positioning data.

2. The vehicle monitoring method based on BeiDou positioning as described in claim 1, characterized in that, The structured BeiDou positioning raw dataset includes pseudorange, carrier phase, Doppler shift, satellite azimuth, satellite ID information, and timestamp information. The step of performing time synchronization correction on the corrected location data to obtain a time-synchronized location dataset includes: The length of the satellite signal propagation path is obtained from the original BeiDou positioning dataset. The satellite signal propagation path length is corrected based on the vehicle's altitude information to obtain the corrected satellite signal propagation path length. The signal propagation time delay information is calculated based on the electromagnetic wave propagation speed and the corrected satellite signal propagation path length. The original timestamp information is corrected based on the signal propagation time delay information to obtain the corrected timestamp information; The corrected timestamp information is correlated with other data in the original BeiDou positioning dataset to obtain the time-synchronized positioning dataset.

3. The vehicle monitoring method based on BeiDou positioning as described in claim 2, characterized in that, The process of matching vehicle location based on vehicle 3D coordinates and an electronic map to obtain vehicle location information includes: Transform the vehicle's three-dimensional coordinates from the geocentric coordinate system to the Gauss-Kruger plane coordinate system. Extract the index information of the environmental feature mapping library from the electronic map; Based on the Gauss-Kruger plane coordinates of the vehicle's three-dimensional coordinates and the index information of the environmental feature mapping library in the electronic map, the spatial point-in-polygon algorithm is used to determine the boundary range of the vehicle's coordinates. The vehicle's three-dimensional coordinates are corrected based on the boundary range to obtain the corrected vehicle three-dimensional coordinates.

4. The vehicle monitoring method based on BeiDou positioning as described in claim 3, characterized in that, The process of matching vehicle location based on vehicle 3D coordinates and an electronic map to obtain vehicle location information includes: Based on the vehicle's three-dimensional coordinates and the electronic map, obtain the vehicle's planar coordinate sequence and spatial data of surrounding candidate roads; The degree of alignment between each candidate road and the trajectory segment is obtained based on the vehicle's planar coordinate sequence and the spatial data of surrounding candidate roads. The integral of the positional deviation between each candidate road and the trajectory segment is calculated based on the vehicle's planar coordinate sequence and the spatial data of the surrounding candidate roads. The specific road identifier of the vehicle is obtained by integrating the degree of alignment between each candidate road and the trajectory segment, as well as the positional deviation between each candidate road and the trajectory segment.

5. The vehicle monitoring method based on BeiDou positioning as described in claim 4, characterized in that, The process of obtaining vehicle trajectory information based on vehicle location information includes: An aligned planar coordinate time series is generated based on the specific road signs where the vehicle is located and the time-synchronized positioning dataset. The trajectory curvature is calculated based on the aligned planar coordinate time series, thereby obtaining the curvature value and timestamp corresponding to each positioning point; The vehicle's trajectory information is generated based on the curvature value and timestamp corresponding to each positioning point.

6. The vehicle monitoring method based on BeiDou positioning as described in claim 5, characterized in that, The method of determining vehicle violations and identifying dangerous driving behaviors based on vehicle trajectory information includes: Obtain real-time environmental data; An environmental impact coefficient is generated based on real-time environmental data. Obtain the violation rules; Information on violations is generated based on violation rules, environmental impact coefficients, and vehicle trajectory information.

7. The vehicle monitoring method based on BeiDou positioning as described in claim 6, characterized in that, The method of determining vehicle violations and identifying dangerous driving behaviors based on vehicle trajectory information further includes: The trajectory spatial distribution entropy value of each trajectory window is obtained based on the trajectory information of the vehicle; Calculate the rate of change of entropy values ​​for the spatial distribution entropy of the trajectory between every two adjacent trajectory windows; Whether dangerous driving behavior exists is determined based on the rate of change of each entropy value.

8. The vehicle monitoring method based on BeiDou positioning as described in claim 7, characterized in that, The trajectory spatial distribution entropy value is calculated using the following formula: ； in, The entropy value represents the spatial distribution of the trajectory. This represents the total number of outermost grid cells; This represents the total number of inner layer grids; The number of trajectory points in the j-th inner grid within the i-th outer grid; This represents the total number of points within the window. Dynamic weights for the inner mesh; This is the window-level adjustment coefficient; This is the minimum value correction factor; This is the spatial attenuation factor of the outer grid.

9. A vehicle monitoring system based on BeiDou positioning, used in the vehicle monitoring method based on BeiDou positioning as described in claim 1, characterized in that, The vehicle monitoring system based on BeiDou positioning includes: A BeiDou positioning raw data set acquisition module is used to acquire the structured BeiDou positioning raw data set transmitted by BeiDou satellites; A quality stratification module is used to perform quality stratification on the acquired BeiDou positioning raw dataset to obtain a positioning dataset with quality level labels, including excellent labels, normal labels and poor labels. The correction module is used to initiate multi-window feature comparison and correction on the data corresponding to ordinary labels and inferior labels, so as to obtain the corrected positioning data. A time synchronization module is used to perform time synchronization correction on the corrected positioning data, thereby obtaining a time-synchronized positioning dataset. A vehicle 3D coordinate acquisition module is used to acquire the vehicle's 3D coordinates based on a time-synchronized positioning dataset. An electronic map acquisition module, wherein the electronic map acquisition module is used to acquire electronic maps; The vehicle location information acquisition module is used to match the vehicle's location with the vehicle's three-dimensional coordinates and an electronic map, thereby acquiring the vehicle's location information. A trajectory information acquisition module, which is used to acquire the trajectory information of a vehicle based on the vehicle's location information; The danger assessment module is used to determine the vehicle's violations and identify dangerous driving behaviors based on the vehicle's trajectory information.

Images