Vehicle abnormal state correction method based on thunder-vision fusion

Abstract

The invention discloses a vehicle abnormal state correction method based on thunder-vision fusion. Comprising the following steps of original vehicle coordinate data point interval judgment, driving direction smoothing processing, lane direction smoothing processing, vehicle urgent acceleration or urgent deceleration state detection, vehicle urgent acceleration or urgent deceleration state coordinate correction, adjacent vehicle position identification on the same lane, relative position deviation judgment between vehicles and vehicle abnormal relative position coordinate correction. According to the invention, a co-processing framework of "single vehicle state correction" and "workshop relationship correction" is constructed, two types of abnormities with different properties are incorporated into a unified process for processing, and the limitation of single processing dimension of an existing method is overcome; according to the method, the abnormal state of the bicycle is corrected firstly, then the relation between the bicycles is judged and corrected based on the more accurate bicycle track, a forward data quality enhancement closed loop is formed, the finally output coordinate data can reflect the running state and the relative position of the bicycle more truly, and the method has wide application prospects and practical value.

Description

Technical Field

[0001] This invention relates to the field of road traffic technology, and in particular to a method for correcting abnormal vehicle states based on radar-visual fusion. Background Technology

[0002] With the rapid development of intelligent transportation technology, radar-video integrated systems have been widely used in road traffic monitoring. Their fixed-period output of longitude and latitude coordinates can be used to reconstruct vehicle trajectories, determine lane affiliation, and provide basic data for traffic monitoring and early warning, as well as autonomous driving decision-making. However, affected by factors such as sensor noise, environmental occlusion, and target matching errors, radar-video coordinate data is prone to short-term fluctuations, jumps, and gaps. Especially under conditions of rapid acceleration and deceleration, there may be significant deviations between the original coordinates and the actual vehicle operating state. Simultaneously, adjacent vehicles in the same lane may exhibit abnormal relative positions, such as a following vehicle "overtaking" the preceding vehicle on the coordinates or an abnormally small distance between vehicles, affecting the reliability of lane-level trajectory analysis and safety risk assessment.

[0003] Existing radar coordinate data processing methods mostly focus on static error correction of single vehicle trajectories. They usually reduce measurement noise by simply filtering or smoothing, lacking in-depth mining and utilization of the time series characteristics of the data. They are insufficient in the comprehensive identification and quantitative correction of abnormal vehicle operation such as rapid acceleration and deceleration, as well as abnormal relative positions of adjacent vehicles in the same lane, making it difficult to restore the true state of the vehicle and the reasonable vehicle-vehicle relationship in a timely and accurate manner.

[0004] Therefore, there is an urgent need for a method that can comprehensively identify and correct abnormal operating states of vehicles such as rapid acceleration and deceleration, as well as abnormal relative positions of vehicles, in order to improve the accuracy and usability of radar coordinate data. Summary of the Invention

[0005] To address the aforementioned issues, this invention provides a vehicle anomaly state correction method based on radar-visual fusion. Utilizing coordinates and lane information output by a radar-visual integrated machine, and through a series of ordered processing steps, it achieves unified modeling, comprehensive identification, and accurate correction of abnormal vehicle operation states such as rapid acceleration and deceleration, as well as abnormal relative positions of vehicles within the same lane. This provides reliable data support for applications such as traffic incident detection and intelligent driving assistance, and solves the problem of insufficient comprehensive processing capabilities of existing technologies for abnormal vehicle operation such as rapid acceleration and deceleration, as well as abnormal relative positions of vehicles within the same lane.

[0006] To address the problems mentioned in the background section and achieve the aforementioned technical objectives, the present invention provides the following technical solution:

[0007] A vehicle abnormal state correction method based on radar-visual fusion includes the following steps: original vehicle coordinate data point interval judgment, driving direction smoothing processing, lane direction smoothing processing, vehicle rapid acceleration or deceleration state detection, vehicle rapid acceleration or deceleration state coordinate correction, adjacent vehicle position identification in the same lane, relative position deviation judgment between vehicles, and abnormal relative position coordinate correction of vehicles.

[0008] Preferably, the original vehicle coordinate data point interval determination step includes: a preset latitude and longitude coordinate interval; if the longitude and latitude of the original vehicle coordinate data both fall within the coordinate interval, it is determined to be valid coordinate data and retained; if any coordinate component exceeds the coordinate interval, it is determined to be invalid coordinate data and discarded.

[0009] Preferably, the driving direction smoothing process includes: calculating the arithmetic mean of the valid coordinate data in the longitude and latitude directions for a time interval of 1 second, and using the obtained mean as the final corrected coordinate for the time interval; if some data points are missing or are determined to be invalid coordinate data in the time interval, then only the valid coordinate data that actually exist in the time interval are averaged.

[0010] Preferably, the lane smoothing process includes: based on the effective coordinate data, using the least squares method to perform polynomial fitting on the latitude and longitude coordinates of each lane centerline, establishing a functional relationship between longitude and latitude, and obtaining the longitude-latitude fitting formula for each lane centerline; determining the lane to which the data point belongs through the lane_id field of the original vehicle coordinate data, selecting the longitude-latitude fitting formula of the corresponding lane centerline and substituting it into the latitude coordinates after lane smoothing, and calculating the corrected longitude coordinates for lane smoothing.

[0011] Preferably, the vehicle rapid acceleration or rapid deceleration state detection step includes: calculating the vehicle's forward distance in the latitudinal direction and comparing it with preset acceleration and deceleration thresholds; if the forward distance is ≥ the preset acceleration threshold, then the vehicle is determined to be in a rapid acceleration state; if the forward distance is ≤ the preset deceleration threshold, then the vehicle is determined to be in a rapid deceleration state.

[0012] Preferably, the vehicle rapid acceleration or rapid deceleration state coordinate correction step includes: standardizing and correcting the vehicle's forward distance under different driving conditions; using the corresponding corrected forward distance to correct the latitude coordinates of the previous moment to obtain the corrected latitude of the current moment; and substituting the corrected latitude into the longitude-latitude fitting formula of the corresponding lane centerline to calculate the corrected longitude coordinates of the current moment.

[0013] Preferably, the formula for calculating the corrected latitude at the current moment is:

[0014] The corrected latitude at the current moment = the latitude coordinates of the previous moment + the corrected forward distance * 0.00000899.

[0015] Preferably, the adjacent vehicle position identification step in the same lane includes: for the driving status at each time point, acquiring the trajectory data records of the two vehicles at that time point, and extracting the lane_id field from them; comparing whether the lane_id field values ​​of the two vehicles at that time point are exactly the same; if they are the same, they are determined to be in the same lane, and the subsequent process continues; if they are not the same or if any vehicle's lane_id field is missing, invalid, or has other abnormalities, the data processing at the current time point is terminated, and the data processing at the next time point is started.

[0016] Preferably, the relative position deviation judgment step between vehicles includes: combining the vehicle driving direction, distinguishing the front vehicle and the rear vehicle by comparing the latitude of the two vehicles at the previous moment; calculating the latitude difference between the two vehicles at the next moment; if the latitude difference is less than or equal to a preset latitude difference threshold, then the relative position of the two vehicles is determined to be abnormal and the subsequent abnormal handling process is triggered; otherwise, the process is terminated.

[0017] Preferably, the abnormal relative position coordinate correction step includes: using the coordinates of the preceding vehicle at the current moment as a reference, correcting the latitude coordinates of the following vehicle: corrected latitude coordinates of the following vehicle = latitude coordinates of the preceding vehicle - preset latitude difference threshold; according to the lane centerline fitting formula of the lane where the following vehicle is located, calculating the corrected longitude coordinates of the following vehicle from the corrected latitude coordinates, and obtaining the corrected coordinates of the abnormal relative position of the vehicle.

[0018] The beneficial effects of this invention are as follows: This invention constructs a collaborative processing framework for "single vehicle state correction" and "vehicle-to-vehicle relationship correction," incorporating two different types of anomalies (self-dynamic anomalies and spatial relationship anomalies) into a unified process for processing, overcoming the limitation of existing methods that only handle a single dimension; by first correcting the abnormal state of a single vehicle, and then judging and correcting the vehicle-to-vehicle relationship based on a more accurate single vehicle trajectory, a positive data quality enhancement closed loop is formed, enabling the final output coordinate data to more realistically reflect the vehicle's operating state and relative position; the method has a clear process, and each step can be parameterized according to specific scenarios (such as various thresholds), making it easy to deploy and optimize in actual traffic data processing systems; at the same time, its framework also reserves space for integrating data from other sensors or more complex anomaly models; the corrected high-precision, high-reliability vehicle trajectory coordinates can directly serve high-precision traffic event detection, lane-level traffic flow analysis, vehicle trajectory prediction, and intelligent connected vehicle decision-making, improving the overall efficiency of intelligent transportation systems. Attached Figure Description

[0019] Figure 1 This is a flowchart of the method of the present invention.

[0020] Figure 2 This is a schematic diagram of the lane centerline during the lane smoothing process.

[0021] Figure 3 In the step of determining the relative positional deviation between vehicles, t=t n-1 A diagram showing the positions of adjacent vehicles at the time.

[0022] Figure 4 In the step of determining the relative positional deviation between vehicles, t=t n A diagram showing the positions of adjacent vehicles at the time. Detailed Implementation

[0023] To enable those skilled in the art to better understand the technical solutions of the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments in this specification. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0024] This invention provides a vehicle abnormal state correction method based on radar-visual fusion. The method involves filtering and smoothing the original vehicle coordinate data output by the radar-visual fusion system to obtain preprocessed coordinates. Based on the preprocessed coordinates, the method detects the rapid acceleration or deceleration of vehicles and corrects the coordinates of vehicles in abnormal states to obtain single-vehicle state-corrected coordinates. Based on the single-vehicle state-corrected coordinates, the method identifies adjacent vehicle pairs in the same lane, determines whether there is a relative positional deviation between the vehicle pairs, and corrects the relative position of the vehicles with deviations.

[0025] like Figure 1 As shown, the vehicle abnormal state correction method based on radar-visual fusion of the present invention specifically includes the following steps:

[0026] S1. Determining the range of original vehicle coordinate data points

[0027] Determining the range of original vehicle coordinate data points is the first step in the original vehicle coordinate data processing flow. For the latitude and longitude coordinates output by the radar-visual fusion system / radar-visual integrated machine with a 100-millisecond cycle, it is necessary to pre-set an appropriate latitude and longitude coordinate range based on the actual geographical information of the target road segment. The specific implementation steps are as follows:

[0028] S11: Specific rules for presetting a reasonable coordinate range.

[0029] Collect the latitude and longitude coordinates of key points such as the start and end points and curvature change points of all lane lines on the target road segment. Calculate the maximum and minimum values ​​of longitude and latitude, and then extend these values ​​along both the longitude and latitude directions with a certain tolerance. The specific tolerance value is selected based on factors such as road type, forming an effective data boundary. This boundary range is the preset reasonable coordinate range, which can be represented in the following form:

[0030] Valid longitude range: [X_min, X_max];

[0031] Valid latitude range: [Y_min, Y_max].

[0032] S12: Data point judgment and filtering processing.

[0033] Iterate through the latitude and longitude data points input every 100 milliseconds. If both the longitude and latitude fall within the above interval, the data point is considered valid and retained. If either coordinate component exceeds the interval, the data point is marked as invalid and discarded, and will not proceed to the next processing step.

[0034] S2, Driving Direction Smoothing Processing

[0035] The smoothing process uses 1 second as the basic time unit to smooth the valid latitude and longitude coordinate data obtained in step S1. Specifically, within each 1-second time interval, the radar-visual fusion system acquires multiple sets of coordinate values. For all coordinate points within the current second that are not determined to be invalid or missing, their arithmetic mean is calculated in both the X and Y directions, and the resulting mean is used as the final corrected coordinate for that time interval. If some sampling points are missing within a certain time interval, only the valid coordinate points that actually exist within that second are averaged. The specific calculation steps are as follows:

[0036] S21: When calculating the X coordinate (lng - longitude), the time interval is 1 second. All valid longitude coordinate values ​​within each 1 second are summarized and their arithmetic mean is calculated. This average value is used as the corrected longitude coordinate after smoothing the direction of travel for that second. The specific formula is as follows:

[0037]

[0038] In the formula, For the radar camera in t n Longitude is measured at all times. For t n Longitude after time correction.

[0039] Note: Since the Rave Integrated Machine outputs a coordinate every 100 milliseconds, n=10 is usually used here, taking into account the case of missing data.

[0040] S22: For the Y-coordinate (lat-latitude), the same calculation steps as in S21 are used. Within each 1-second time interval, all valid latitude coordinate values ​​within each second are summarized (missing or invalid data are removed), and the arithmetic mean is calculated. This arithmetic mean is used as the corrected latitude coordinate after smoothing the heading for that second. The calculation formula is as follows:

[0041]

[0042] In the formula, For the radar camera in t n Longitude is measured at all times. For t n Latitude after time correction.

[0043] S3, Lane Smoothing

[0044] Lane smoothing first uses the least squares method to perform polynomial fitting on the latitude and longitude coordinates of each lane's centerline based on the effective coordinate data, establishing a functional relationship between longitude and latitude. This yields the longitude-latitude fitting formula for each lane's centerline. Then, the lane_id field of the original vehicle coordinate data determines the lane to which the data point belongs, and the corresponding lane's fitting formula is selected. The smoothed Y-coordinate (lat-latitude) is then substituted into the formula to calculate the corrected X-coordinate (lng-longitude) for lane smoothing. The specific lane smoothing processing steps are as follows:

[0045] S31: Based on the lane centerline coordinate data processed in steps S1 and S2, the least squares method is used to fit the longitude and latitude coordinates of the two lane centerlines with longitude as the dependent variable and latitude as the independent variable, and the polynomial fit is performed to obtain the functional relationship between the longitude and latitude of the two lane centerlines, i.e., the K2 fitting formula.

[0046] Based on this, and taking into account the prior knowledge of consistent lane spacing in actual road design, the fitting formulas for the centerline K1 of one lane and the centerline K3 of three lanes are derived by means of translation offset. Finally, the longitude-latitude fitting formulas for the centerlines of the three lanes are obtained, as shown in the diagram. Figure 2 As shown.

[0047] S32: Based on the lane_id field in the radar detection results, the range of X coordinate (lng-longitude) is divided to determine the lane to which the data point belongs and match the correction formula for the corresponding lane. The specific correspondence is shown in Table 1 below.

[0048] Table 1: Lane Correction Formulas Corresponding to the lane_id Field

[0049] lane_id X-coordinate (lng - longitude) corresponding formula 0 1 2

[0050] S33: Substitute the smoothed Y coordinate (lat-latitude) into the fitting formula (K1, K2, K3) of the corresponding lane centerline to calculate the corrected X coordinate (lng-longitude) for lane smoothing.

[0051] S4. Vehicle rapid acceleration or deceleration detection

[0052] In this step, to identify whether the vehicle is in a state of rapid acceleration or deceleration and to provide a basis for subsequent coordinate correction, the changes in the latitudinal coordinates of the same vehicle at adjacent time points are analyzed. By calculating the vehicle's travel distance in the latitudinal direction and comparing this distance with preset acceleration and deceleration thresholds, the corresponding driving state of the vehicle is determined. The specific detection steps are as follows:

[0053] S41: Calculate the distance the vehicle has traveled.

[0054] Based on the continuous latitudinal coordinate data of the vehicle obtained through steps S1-S3, the latitudinal difference between adjacent moments is calculated, and the conversion relationship between Earth's latitude and longitude and actual distance is combined to obtain the actual distance traveled by the vehicle in the latitudinal direction, which is the vehicle's forward distance. The specific calculation formula is as follows:

[0055]

[0056] In the formula, Let be the distance traveled at time n. Let be the latitude coordinates at time n. Let be the latitude coordinates at time n-1.

[0057] S42: By comparing the calculated vehicle forward distance with a set threshold, determine whether the vehicle is in a special driving state of rapid acceleration or rapid deceleration. The specific vehicle state judgment method is shown in Table 2 below.

[0058] Table 2: Vehicle Status Assessment

[0059] forward distance Vehicle status assessment rapid acceleration rapid deceleration

[0060] Note: Y_d n Let α be the distance traveled at time n; α and β are the thresholds for judging the rapid acceleration and deceleration of the vehicle, and their specific values ​​depend on factors such as the relative distance between vehicles, road conditions, and the vehicle's own performance.

[0061] S5. Vehicle acceleration or deceleration state coordinate correction

[0062] Based on the detection results of the aforementioned rapid acceleration or deceleration of vehicles, for vehicles determined to be in abnormal driving states, further coordinate correction is required to ensure that the trajectory coordinates recorded under these states more accurately reflect the vehicle's actual spatial position on the road. Therefore, this step standardizes the vehicle's forward distance under different driving states and updates the latitude and longitude coordinates at the current moment accordingly, thus obtaining the state-corrected vehicle position. The specific coordinate correction steps are as follows:

[0063] S51: The vehicle's forward distance is standardized and corrected according to different vehicle conditions. The specific correction method is shown in Table 3 below.

[0064] Table 3: Corrected Forward Distance

[0065] Vehicle status Correct forward distance rapid acceleration rapid deceleration

[0066] Note: Y_dl n γ represents the corrected forward distance at time n; γ and θ represent the corrected forward distance under conditions of rapid acceleration and deceleration of the vehicle, and their specific values ​​depend on factors such as the relative distance between vehicles, road conditions, and the vehicle's own performance.

[0067] S52: Based on the vehicle's rapid acceleration or deceleration, the latitude coordinates of the previous moment are corrected using the corresponding corrected forward distance to obtain the corrected latitude of the current moment. The specific calculation formula is as follows:

[0068]

[0069] In the formula, Let be the latitude coordinates at time n-1; The corrected latitude coordinates at time n; This is for correcting the forward distance under conditions of rapid acceleration and rapid deceleration.

[0070] S53: Based on the relationship of the lane_id field values ​​in Table 1 of step S32, select the fitting formula (K1, K2, K3) for the lane centerline, substitute the above-mentioned corrected latitude into the fitting formula for the lane centerline, and thus calculate the corrected longitude coordinates at the current time.

[0071] S6, Adjacent vehicle position recognition in the same lane

[0072] After completing the coordinate correction for the aforementioned rapid acceleration or deceleration of vehicles, in order to further identify anomalies in the relative positions between vehicles, it is necessary to first filter out adjacent vehicle pairs in the same lane from the traffic flow to provide objects for subsequent judgment of relative position deviations between vehicles. The specific implementation is as follows:

[0073] For each second, the radar-visual fusion system acquires the trajectory data records of vehicle A and vehicle B for that second and extracts the lane_id field from them. By comparing whether the lane_id field values ​​of the two vehicles for that second are exactly the same, it determines whether they are in the same lane. If the lane_ids are the same, they are determined to be in the same lane, and the subsequent analysis steps continue. If they are not the same, or if any vehicle's lane_id field is missing, invalid, or has other abnormalities, the data processing flow for the current second is terminated, and the system moves on to process the data for the next time point.

[0074] S7. Judgment of relative position deviation between vehicles

[0075] For identified pairs of adjacent vehicles in the same lane, to determine whether there are abnormally close proximity or positional inversions between the two vehicles, it is necessary to analyze the distance between them by combining the changes in latitudinal coordinates over consecutive time intervals. By comparing the latitudinal difference between the preceding and following vehicles at adjacent time intervals with a preset threshold, it can be determined whether the relative positions between the vehicles are within a normal range. The specific judgment steps are as follows:

[0076] S71: By comparing the previous time step (i.e., t=t) n-1 Based on the Y-coordinates (lat-latitude) of vehicles A and B, and their respective directions of travel, B is determined to be the preceding vehicle and A to be the following vehicle. A diagram illustrating the vehicle positions is shown below. Figure 3 As shown.

[0077] S72: Calculate the next time step (i.e., t=t) n (At time) The latitude difference L between vehicle A and vehicle B: L=Y 前 -Y 后 The diagram is as follows Figure 4 As shown.

[0078] S73: Determine the latitude difference threshold ε based on factors such as sensor accuracy, traffic scenario, lane width, and vehicle size; if the latitude difference L ≤ latitude difference threshold ε, then determine that the relative position of the two vehicles is abnormal and trigger the subsequent abnormal handling process, i.e., step S8; otherwise, terminate the process.

[0079] S8, Correction of abnormal relative position coordinates of vehicles

[0080] If step S7 determines that there is an anomaly in the relative position between adjacent vehicles, the coordinates of the following vehicle need to be corrected so that the corrected vehicle coordinates can more accurately reflect the actual relative position of the vehicles on the road. This step uses the coordinates of the current preceding vehicle as a reference, adjusts the latitude coordinates of the following vehicle, and updates its longitude coordinates based on the lane centerline fitting relationship, thereby obtaining the corrected coordinates for the abnormal relative position of the vehicles. The specific implementation steps are as follows:

[0081] S81: Based on the judgment and data in step S7, when t=t n At any given moment, the coordinates of the preceding vehicle B remain unchanged. Instead, the corrected latitude coordinates of the following vehicle A are obtained by subtracting a correction factor, the latitude difference threshold β, from the Y-coordinate (lat-latitude) of the preceding vehicle B. This correction is denoted as [missing information]. The specific calculation formula is as follows:

[0082]

[0083] In the formula, Corrected latitude coordinates for the following vehicle. The latitude coordinates of the vehicle in front. The latitude difference threshold is determined based on factors such as sensor accuracy, traffic scenario, lane width, and vehicle size.

[0084] S82: Based on the rear vehicle A at t=t n Based on the lane_id field recorded at each moment, and according to the relationship between the lane_id field values ​​in Table 1 of step S32, the fitting formula (K1, K2, K3) for the lane centerline is selected, thereby obtaining the new latitude coordinates of the following vehicle A. Substituting the values ​​into the lane centerline fitting formula for the lane where vehicle A is currently located, the corrected longitude coordinates of vehicle A can be calculated, thus obtaining the corrected coordinates of vehicle A. express.

[0085] In summary, the vehicle abnormal state correction method based on radar-visual fusion of the present invention has the following advantages over the prior art:

[0086] (1) This invention breaks through the limitation of existing methods that only focus on the static error correction of single vehicle trajectory. It makes full use of the time series characteristics of radar coordinate data and lane spatial relationship to construct a collaborative processing framework for single vehicle abnormal state correction and lane relationship correction. It realizes the collaborative correction of abnormal running states of vehicles such as rapid acceleration and deceleration and abnormal relative positions of vehicles in the same lane, effectively solving the problem of insufficient comprehensive processing capability of multiple abnormal states in the existing technology.

[0087] (2) By using the original data point interval judgment step, valid coordinate data can be quickly filtered out and invalid data can be eliminated, providing a high-quality data foundation for subsequent processing and improving the efficiency and accuracy of overall data processing.

[0088] (3) The combination of driving direction smoothing and lane direction smoothing optimizes the coordinate data from the time dimension and the spatial dimension respectively, effectively reducing the coordinate fluctuations caused by sensor noise, environmental occlusion and other factors, and improving the stability and reliability of the coordinate data.

[0089] (4) Scientific and reasonable detection mechanisms and correction algorithms were designed for the rapid acceleration and deceleration of vehicles and the abnormal relative positions of vehicles in the same lane. These mechanisms can accurately identify abnormal states and make quantitative corrections. Based on more accurate single-vehicle trajectories, the relationship between vehicles and lanes is judged and corrected, forming a positive data quality enhancement closed loop. This makes the corrected coordinate data more realistically reflect the actual operating status of vehicles and the positional relationship within the lane.

[0090] (5) The method of the present invention has a clear process and strong operability. The corrected radar coordinate data can be directly used for road traffic monitoring, lane-level trajectory analysis and intelligent driving assistance, etc., providing strong data support for the efficient operation of intelligent transportation systems. It has broad application prospects and practical value.

[0091] Regarding the embodiments of the present invention, it should also be noted that, without conflict, the embodiments of the present invention and the features in the embodiments can be combined with each other to obtain new embodiments.

[0092] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. The scope of protection of the present invention should be determined by the scope of the claims. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for correcting abnormal vehicle states based on radar-visual fusion, characterized in that, Includes the following steps: The system includes: original vehicle coordinate data point interval judgment, driving direction smoothing, lane direction smoothing, vehicle rapid acceleration or deceleration state detection, vehicle rapid acceleration or deceleration state coordinate correction, adjacent vehicle position identification in the same lane, relative position deviation judgment between vehicles, and abnormal relative position coordinate correction of vehicles.

2. The method according to claim 1, characterized in that, The original vehicle coordinate data point interval determination step includes: a preset latitude and longitude coordinate interval; if the longitude and latitude of the original vehicle coordinate data both fall within the coordinate interval, it is determined to be valid coordinate data and retained; if any coordinate component exceeds the coordinate interval, it is determined to be invalid coordinate data and discarded.

3. The method according to claim 2, characterized in that, The driving direction smoothing process includes: calculating the arithmetic mean of the valid coordinate data in the longitude and latitude directions for a time interval of 1 second, and using the obtained mean as the final corrected coordinate for the time interval; if some data points are missing or are determined to be invalid coordinate data in the time interval, then only the valid coordinate data that actually exist in the time interval are averaged.

4. The method according to claim 3, characterized in that, The lane smoothing process includes: based on the valid coordinate data, using the least squares method to perform polynomial fitting on the latitude and longitude coordinates of each lane centerline, establishing a functional relationship between longitude and latitude, and obtaining the longitude-latitude fitting formula for each lane centerline; determining the lane to which the data point belongs through the lane_id field of the original vehicle coordinate data, selecting the longitude-latitude fitting formula of the corresponding lane centerline and substituting it into the latitude coordinates after lane smoothing, and calculating the corrected longitude coordinates for lane smoothing.

5. The method according to claim 4, characterized in that, The vehicle rapid acceleration or rapid deceleration state detection step includes: calculating the vehicle's forward distance in the latitudinal direction and comparing it with preset acceleration and deceleration thresholds; if the forward distance is ≥ the preset acceleration threshold, the vehicle is determined to be in a rapid acceleration state; if the forward distance is ≤ the preset deceleration threshold, the vehicle is determined to be in a rapid deceleration state.

6. The method according to claim 5, characterized in that, The vehicle acceleration or deceleration state coordinate correction step includes: standardizing and correcting the vehicle's forward distance under different driving conditions; using the corresponding corrected forward distance to correct the latitude coordinates of the previous moment to obtain the corrected latitude of the current moment; and substituting the corrected latitude into the longitude-latitude fitting formula of the corresponding lane centerline to calculate the corrected longitude coordinates of the current moment.

7. The method according to claim 6, characterized in that, The formula for calculating the corrected latitude at the current moment is: The corrected latitude at the current moment = the latitude coordinates of the previous moment + the corrected forward distance * 0.00000899.

8. The method according to claim 6, characterized in that, The adjacent vehicle location identification step in the same lane includes: for the driving status at each time point, acquiring the trajectory data records of the two vehicles at that time point, and extracting the lane_id field from them; comparing whether the lane_id field values ​​of the two vehicles at that time point are exactly the same. If they are the same, they are determined to be in the same lane, and the subsequent process continues; if they are not the same or if any vehicle's lane_id field is missing, invalid, or has other abnormalities, the data processing at the current time point is terminated, and the data processing at the next time point is started.

9. The method according to claim 8, characterized in that, The steps for determining the relative position deviation between vehicles include: combining the vehicle's driving direction, distinguishing between the front and rear vehicles by comparing the latitude of the two vehicles at the previous moment; calculating the latitude difference between the two vehicles at the next moment; if the latitude difference is less than or equal to a preset latitude difference threshold, then the relative position of the two vehicles is determined to be abnormal and subsequent abnormal handling procedures are triggered; otherwise, the process is terminated.

10. The method according to claim 9, characterized in that, The abnormal relative position coordinate correction step of the vehicle includes: using the coordinates of the preceding vehicle at the current moment as a reference, correcting the latitude coordinates of the following vehicle: corrected latitude coordinates of the following vehicle = latitude coordinates of the preceding vehicle - preset latitude difference threshold; according to the lane centerline fitting formula of the lane where the following vehicle is located, calculating the corrected longitude coordinates of the following vehicle from the corrected latitude coordinates, and obtaining the corrected coordinates of the abnormal relative position of the vehicle.

Images