Vehicle trajectory data adaptive processing method based on vehicle-cloud integration

Abstract

The application provides a vehicle track data adaptive processing method of vehicle cloud integration, comprising the following steps: S1, data acquisition, using multiple data acquisition devices to collect vehicle track data at the vehicle end, the collected vehicle track data comprising the spatial coordinates and time stamp of the vehicle, and transmitting the collected vehicle track data to the computing device of the vehicle through the data bus or wireless communication mode in the vehicle; S2, data preprocessing, comprising the sub-steps of redundancy merging, null filling and the like; S3, data correlation, comprising the sub-steps of spatial correlation, data slicing, outlier correction and the like. The application uses multiple means to acquire vehicle track data, effectively avoids the absence of vehicle track data, and improves the real-time performance of the vehicle track data. Through the technical means of outlier correction, redundancy merging and the like, the application improves the rationality of the vehicle track data, the accuracy of track prediction, the efficiency of vehicle track data processing and the real-time performance of feedback.

Description

Technical Field

[0001] This invention belongs to the field of intelligent transportation, specifically relating to an adaptive processing method for vehicle trajectory data that integrates vehicle and cloud computing. Background Technology

[0002] Patent application number 202311582372.5 discloses a multimodal analysis method for trajectory data of autonomous vehicles. The basic technical content of the method includes: obtaining the instantaneous linear quantities and instantaneous angular quantities of each vehicle, wherein the instantaneous linear quantities include instantaneous coordinate position, instantaneous driving speed, and instantaneous driving acceleration, and the instantaneous angular quantities include instantaneous azimuth angle, instantaneous angular velocity, and instantaneous angular acceleration; establishing Gaussian distribution functions for the instantaneous linear quantities and instantaneous angular quantities, and determining the motion mode of each vehicle based on the relationship between the mean and variance of the data in the Gaussian distribution functions of the instantaneous linear quantities and instantaneous angular quantities.

[0003] The invention patent application with application number 202310331776.0 discloses a method for screening abnormal vehicle data based on trajectory data. The method includes the following main steps: (1) obtaining a highway vehicle trajectory dataset and preprocessing the vehicle trajectory dataset; (2) judging the preprocessed vehicle trajectory dataset. If the sampling time interval is lower than a preset threshold, proceed to step (3); otherwise, proceed to step (4); (3) resampling the vehicle trajectory dataset; (4) dividing the vehicle trajectory dataset into sub-datasets according to time and space characteristics; (5) analyzing the data characteristics of vehicles in each sub-dataset and marking abnormal index data as abnormal data based on the analysis results of the data characteristics.

[0004] Each of the aforementioned inventions has its advantages. Specifically, the technical solution described in patent application number 202311582372.5 deeply considers the characteristics and features of autonomous vehicle motion, and constructs a multimodal analysis and verification method for autonomous driving trajectory data. This method can classify vehicle motion behavior under different conditions, providing accurate and reliable data support for key technologies such as path planning and control decisions in autonomous driving systems. The technical solution described in patent application number 202310331776.0 divides trajectory data into subsets based on temporal and spatial features. Within these subsets, feature analysis is performed on three indicators: vehicle speed, acceleration, and headway. This reduces spatiotemporal interference, thereby effectively filtering out abnormal data that occurs during vehicle operation.

[0005] However, the above inventions also have some technical defects, which are mainly manifested in the following two aspects:

[0006] First, the methods for acquiring vehicle trajectory data are relatively limited, leading to missing data or poor real-time performance. Missing data affects the completeness and accuracy of vehicle trajectories. In particular, the aforementioned invention ignores outliers that often exist in vehicle trajectory data, thus affecting the rationality of the data and the accuracy of trajectory prediction. Furthermore, the aforementioned invention lacks in-depth data analysis for special operating conditions (such as vehicle steering) and lacks adaptive processing methods, severely impacting data quality under these special conditions.

[0007] Second, the aforementioned invention generates a large amount of redundant data during the data processing process. This large amount of redundant data increases the burden of transmission and processing, resulting in insufficient real-time data processing and feedback, which seriously affects the timeliness of autonomous driving decisions. Summary of the Invention

[0008] The purpose of this invention is to overcome the first and second technical defects of the prior art.

[0009] To achieve the above objectives, the present invention employs the following technical solution:

[0010] An adaptive processing method for vehicle trajectory data integrating vehicle and cloud computing includes the following steps:

[0011] S1, Data Acquisition; On the vehicle side, data acquisition equipment is used to collect vehicle trajectory data. The collected vehicle trajectory data includes the vehicle's spatial coordinates and timestamps, and the collected vehicle trajectory data is transmitted to the vehicle's computing device through the vehicle's internal data bus or wireless communication.

[0012] S2, Data Preprocessing; The vehicle's computing equipment performs preliminary processing on the vehicle trajectory data, including the following sub-steps:

[0013] S201, Redundancy merging; Set a time window for redundant data and merge vehicle trajectory points with a time difference less than that time window.

[0014] S202, Null value filling; using adaptive window technology to dynamically adjust the sampling window according to the density of vehicle trajectory data;

[0015] S203, Data Upload: Upload the pre-processed vehicle trajectory data to the cloud via wireless network;

[0016] S3, Data Association; includes the following sub-steps:

[0017] S301, Spatial Association; Using geospatial indexes to achieve spatial data retrieval and management, spatially association is performed between vehicle trajectory points after redundancy merging and null value filling and road network data, traffic element information and regional POI information.

[0018] S302, Data Slicing; Slicing geospatial data according to a preset grid size to achieve gridded management of spatial data;

[0019] S303, Outlier Correction: An adaptive outlier filtering algorithm based on a combination of sample entropy and local outlier factor is used to identify outliers in vehicle trajectories.

[0020] S304, Vehicle trajectory information generation; merge the corrected vehicle trajectory points with the normal point set to generate vehicle trajectory information.

[0021] Based on the above technical solutions, the present invention may be supplemented with the following technical means to better or more specifically solve the technical problems to be solved by the present invention:

[0022] The data acquisition equipment includes video surveillance, inductive loops, radio frequency identification (RFID), automotive electronic identification, as well as on-board diagnostic systems and global positioning systems installed in intelligent vehicles.

[0023] Furthermore, when executing S201, the vehicle trajectory data is set as: {(x i ,y i ,t i )},x i and y i t represents the spatial coordinates of the vehicle. i Representing timestamps, a time window Δt is set for redundant data. Vehicle trajectory data points with a time difference less than Δt are merged, and the position of the merged vehicle trajectory point is the weighted average of all vehicle trajectory data points within the time window Δt.

[0024] , , ;

[0025] In the formula, w i It is the weight, w i The value can be set to 1 or according to the reliability of the data points.

[0026] Furthermore, when S202 is executed under special working conditions (such as vehicle steering), encrypted data points are used. Based on the concept of entropy, the complexity of vehicle trajectory data is calculated. For two-dimensional trajectory coordinates, vector difference is used to perform sample entropy analysis to describe the complexity of instantaneous changes in vehicle trajectory, capture nonlinear features in the signal, and set a threshold to adaptively and dynamically adjust the window size and fill null values ​​by judging the degree of change of sample entropy within a certain range.

[0027] Furthermore, executing S202 includes the following steps:

[0028] First, set the initial size to L.S The sampling window is slid towards the vehicle trajectory sequence and the sample entropy of the vehicle trajectory vector in each segment window is calculated.

[0029] Secondly, a threshold is set, and the sample entropy of adjacent segments is subtracted pairwise; segments whose subtraction results exceed the threshold are considered segments with drastic changes in vehicle trajectory curvature.

[0030] Next, obtain the maximum length L that satisfies the threshold judgment condition. max And calculate 1-L respectively max The mean sample entropy corresponding to the vehicle trajectory;

[0031] Next, sort the sample entropy mean from smallest to largest, draw a scatter plot, and connect the minimum and maximum values ​​to form a straight line;

[0032] Finally, calculate the distance l from each point to the line in turn, and the x-coordinate corresponding to the maximum distance lmax is the optimal window length L.

[0033] Furthermore, for a given vehicle trajectory sequence B = {b1, b2, K, b} of length N, n The formula for calculating sample entropy is as follows:

[0034] ;

[0035] In the formula, integers 0 ≤ m ≤ N represent the length of the sequences to be compared, real number r > 0 represents the tolerance for accepting a match, and C m (r) represents the empirical probability that two sequences match at m points, which is obtained by the following formula:

[0036] ,

[0037] In the formula, C i m (r) is obtained through the following formula:

[0038] ;

[0039] In the formula, C i m (r) represents any vector x m (j) in x m The probability of (i) within the range of r. NumT ( x m (i) , x m (j) ) represents x m (i) and x m The number of times that the distance relation (j) is less than r, x m(i) and x m (j) is an m-dimensional vector reconstructed from B.

[0040] Furthermore, linear interpolation is used to fill in the missing values. Let the two known vehicle trajectory points be... and The missing point in the middle is ,but:

[0041] , ;

[0042] After obtaining L, the time window slides by a certain step size, and adaptive processing of the next trajectory sequence begins.

[0043] Furthermore, when executing S302, the entire geographic space is divided into × The grid, the size of each grid × Vehicle trajectory points The network index to which it belongs is:

[0044] ; .

[0045] Furthermore, when executing S303, the outlier degree of the data point in the local region is adaptively calculated based on the density of neighboring points around the data point, including the following steps:

[0046] First, find the k nearest neighbors of the target point d and calculate their reachability distance RD. The reachability distance RD is defined as follows:

[0047] ,

[0048] In the formula, k dis Let d be the distance from point d to its k-th nearest neighbor; dis (d, dj) represents the distance from point d to dj. j The distance (j∈{1,2, …, k});

[0049] Secondly, traverse the vehicle trajectory and calculate the reciprocal of the mean (LRD) of the RD at each point in turn, and store it in set R1;

[0050] Next, the trajectory is traversed again, and the LOF of each point is calculated using R1 and stored in set L1, resulting in:

[0051] ,

[0052] Finally, L1 is judged and filtered based on the threshold c to obtain the final target outlier set.

[0053] Furthermore, the set of vehicle trajectory points is defined as {(x i , y i , t i The outlier is (x) out , y out , t out If the initial outlier is 0, then the final target outlier can be obtained by correcting the two outliers, as shown in the following formula:

[0054] , .

[0055] Furthermore, the cloud analyzes and models historical trajectory data to generate predicted trajectories.

[0056] Compared with the prior art, the present invention has the following beneficial effects:

[0057] First, on the vehicle side, this invention uses multiple methods to acquire vehicle trajectory data, which not only effectively avoids the loss of vehicle trajectory data, but also improves the real-time performance of vehicle trajectory data.

[0058] Secondly, in terms of vehicle trajectory data processing, we attach great importance to outliers in vehicle trajectory data. By adopting outlier correction and other technical means, we have improved the rationality of vehicle trajectory data and the accuracy of trajectory prediction.

[0059] Third, by adopting redundancy merging techniques, the efficiency of vehicle trajectory data processing and the real-time nature of feedback have been improved.

[0060] Fourth, for special working conditions (such as vehicle turning and other driving conditions), encrypted data points are used, and the complexity of vehicle trajectory data is calculated based on the concept of entropy, which improves the quality of vehicle trajectory data under special working conditions. Attached Figure Description

[0061] Figure 1 This is a flowchart of an embodiment of the present invention. Detailed Implementation

[0062] To facilitate a thorough understanding of the technical solution of the present invention by those skilled in the art, an embodiment of the present invention is described below in conjunction with the accompanying drawings.

[0063] like Figure 1 As shown, an adaptive processing method for vehicle trajectory data integrating vehicle and cloud computing includes the following steps:

[0064] S1, Data Acquisition;

[0065] At the vehicle end (referred to as the vehicle-side), a variety of vehicle trajectory data acquisition devices are used, such as surveillance cameras, loop detectors, radio frequency identification (RFID), electronic registration identification (ERI), and on-board diagnostic (OBD) and GPS devices installed in intelligent vehicles to collect vehicle trajectory data. The collected vehicle trajectory data includes the vehicle's spatial coordinates (longitude and latitude) and timestamps. The collected vehicle trajectory data is transmitted to the vehicle's computing device via the vehicle's internal data bus or wireless communication, where it is preprocessed.

[0066] S2, Data Preprocessing;

[0067] Data preprocessing, also known as vehicle trajectory data preprocessing, involves the vehicle's computing equipment performing preprocessing on the received vehicle trajectory data. The purpose of this preprocessing is to reduce the amount of data transmitted. Vehicle trajectory data preprocessing includes the following sub-steps:

[0068] S201, Redundancy merging;

[0069] Set the vehicle trajectory data as {(x i ,y i ,t i )}, where xi and yi represent the spatial coordinates of the vehicle, and t i If the timestamp is used, then for redundant data, a time window Δt can be set to merge vehicle trajectory data points with a time difference less than Δt (merging vehicle trajectory data points corresponds to merging vehicle trajectory points). The position of the merged vehicle trajectory point can be taken as the weighted average of all vehicle trajectory data points within the window, as shown below:

[0070] , , ;

[0071] In the formula, w i It represents the weight, which is usually set to 1 or determined based on the reliability of the data points.

[0072] S202, Null value filling;

[0073] By utilizing adaptive windowing technology, the sampling window is dynamically adjusted based on data density to improve the accuracy of null value filling. It is worth noting that under special operating conditions (such as vehicle steering), a denser data point is needed to obtain more detailed trajectory changes. Therefore, this embodiment calculates the complexity of vehicle trajectory data based on the concept of entropy. For two-dimensional trajectory coordinates, vector difference is used to perform sample entropy analysis to describe the complexity of instantaneous trajectory changes, better capturing the nonlinear characteristics in the signal. Furthermore, by judging the degree of change in sample entropy within a certain range, a threshold is set to adaptively and dynamically adjust the window size and fill null values, enhancing the local change characteristics of the vehicle trajectory, avoiding information loss, and ensuring the continuity and integrity of the vehicle trajectory.

[0074] In this embodiment, the specific steps for filling in null values ​​are as follows:

[0075] First, set the initial size L. S The sampling window slides towards the vehicle trajectory sequence and calculates the sample entropy of the vehicle trajectory vector in each segment window;

[0076] Secondly, a threshold Threshold is set, and the sample entropy of adjacent segments is subtracted pairwise; segments whose subtraction result exceeds the Threshold are regarded as segments with drastic changes in vehicle trajectory curvature.

[0077] Next, obtain the maximum length L that satisfies the threshold judgment condition. max And calculate 1-L respectively max The mean sample entropy corresponding to the vehicle trajectory;

[0078] Next, sort these values ​​from smallest to largest, plot a scatter plot, and connect the minimum and maximum values ​​to form a straight line;

[0079] Finally, calculate the distance l from each point to the line, and the maximum distance l. max The corresponding x-coordinate represents the optimal window length L. This is to ensure that the window covers the entire trajectory under specific conditions, guaranteeing the integrity of short-term trajectory features. For a given vehicle trajectory sequence of length N... The formula for calculating sample entropy is as follows:

[0080] ,

[0081] In the formula, integers 0 ≤ m ≤ N represent the length of the sequences to be compared, real number r > 0 represents the tolerance for accepting a match, and C m (r) represents the empirical probability that two sequences match at m points, which is obtained by the following formula:

[0082] ,

[0083] In the formula, Ci m (r) is obtained through the following formula:

[0084] ;

[0085] In the formula, C i m (r) represents any vector x m (j) in x m The probability of (i) within the range of r. NumT ( x m (i) , x m (j) )express x m (i) and x m (j) The number of times the distance relationship is less than r. x m (i) and x m (j) It is an m-dimensional vector reconstructed from B. When processing the data, linear interpolation is used to fill in the missing values. Let the two known trajectory points be... and The missing point in the middle is ,but:

[0086] , ;

[0087] After obtaining L, the time window slides by a certain step size, and adaptive processing begins on the next segment of the trajectory sequence. Once the entire sequence has been processed, the optimal window length set L can be obtained. A According to L A By dividing the trajectory, complete segments of change information can be obtained.

[0088] S203, Data Upload;

[0089] The vehicle-mounted system uploads the pre-processed vehicle trajectory data to the cloud via a wireless network (such as 4G / 5G). The data transmission volume can be estimated using the following formula:

[0090] ,

[0091] In the formula, D t S represents the total data transmission volume. i For the first The size of each data packet This represents the number of data packets.

[0092] S3, data association;

[0093] Data association includes the following sub-steps:

[0094] S301, Spatial Association;

[0095] By leveraging geospatial indexes (such as R-trees), efficient spatial data retrieval and management can be achieved. This allows for the rapid spatial association of corrected vehicle trajectory points (i.e., vehicle trajectory points after redundancy merging and null value filling) with road network data, traffic element information, and regional POI information. Furthermore, it can accurately locate and associate vehicle trajectory points with other elements in geospatial space (such as roads, traffic signs, and points of interest), enabling refined management and analysis of vehicle travel paths.

[0096] Let the trajectory point of the vehicle be... Spatial data is Find the spatial data closest to the vehicle's trajectory point:

[0097] ;

[0098] By using nearest neighbor search or range query, vehicle trajectory data is correlated with various spatial information, providing a comprehensive spatial data view to support complex traffic management and analysis needs. Geospatial indexing enables precise matching of vehicle trajectory points with road network data and traffic elements, ensuring the spatial accuracy of trajectory data and facilitating a better understanding and management of actual vehicle routes.

[0099] S302, data slice;

[0100] Geospatial data is sliced ​​according to a preset grid size to achieve gridded management of spatial data. This is achieved by dividing the entire space into... × The grid, the size of each grid × This method divides large-scale geospatial data into smaller, more manageable units, significantly improving the efficiency of data processing and querying. Each grid cell can be processed independently, reducing computational complexity.

[0101] trajectory points The network index to which it belongs is:

[0102] , ;

[0103] The grid size can be adjusted according to specific application needs. Smaller grids are used for finer spatial segmentation and management, while larger grids are more suitable for macro-analysis and large-scale data processing. This is applicable to everything from micro-level traffic management to macro-level urban planning. Furthermore, grid slicing facilitates real-time updates and corrections of vehicle trajectory data. New data can be quickly located in the corresponding grid cells for updates, improving the data's dynamic adaptability.

[0104] S303, Outlier Correction;

[0105] An adaptive outlier removal algorithm based on a combination of sample entropy and local outlier factors is used to identify outliers in the trajectory. Outliers are identified and corrected based on speed, time, distance, and current position to ensure the entire trajectory conforms to kinematic laws. The adaptive removal technique improves the sensitivity and accuracy of outlier detection and correction. The initial window size is set to L. S The algorithm performs sliding detection on outliers in the trajectory with minimal permissible computational cost. The sliding distance is T, and a threshold is set based on the deviation characteristics of the outliers. The window first slides across the entire trajectory sequence with an initial size and a distance of T, locating and extracting segments (ST) that exceed the threshold. Based on ST, the window slides again. At this point, through the adaptive adjustment principle of the window size, preliminary accurate detection of outliers can be achieved. Subsequently, using the local outlier factor algorithm, based on the preliminary screening results, and considering the relationship between the data point and the overall data, the algorithm adaptively calculates the outlier degree of the data point in the local region according to the density of its neighboring points. The main steps are as follows:

[0106] First, find the k nearest neighbors of the target point d and calculate their reachability distance RD. The reachability distance RD is defined as follows:

[0107] ,

[0108] In the formula, k dis Let d be the distance from point d to its k-th nearest neighbor; dis (d, d j () is the distance from point d to d j The distance, j∈{1, 2,…, k}.

[0109] Secondly, the vehicle trajectory is traversed and the reciprocal of the mean (LRD) of each point is calculated and stored in set R1. Next, the trajectory is traversed again and the LOF (Local Outlier Factor) of each point is calculated using R1 and stored in set L1. The LOF corresponding to point d is as follows:

[0110] ,

[0111] Finally, L1 is judged and filtered based on the threshold c to obtain the final target outlier set. The vehicle trajectory point set is set as {(x... i ,y i ,t i The outlier is (x) out ,y out ,t out Then, the final target outlier can be obtained by correcting the two outliers:

[0112] , .

[0113] S304, Trajectory information generation;

[0114] The vehicle trajectory points corrected in sub-step S303 are merged with the normal point set, and the generated trajectory information can be represented as {(x j ,y j ,attributes j )}, where attributes j This includes associated road network data, traffic element information, and regional POI information. The cloud merges the corrected vehicle trajectory points with the normal point set to generate complete vehicle trajectory information, and synchronizes the results back to the vehicle for real-time updates and feedback. Incremental transmission technology is used, and the data transmission formula is:

[0115] ,

[0116] In the formula, T incremental Let t1 be the data transmission time, D be the total data transmission amount, t1 and t2 be the target synchronization time, and size(d) be the size of a single data record. After receiving ΔD(t1,t2), the receiver updates its local dataset D(t1) to the latest state D(t2), and the data relationship is as follows:

[0117] ;

[0118] In the formula, D local This is a local dataset.

[0119] Meanwhile, the cloud analyzes and models historical trajectory data to generate predicted trajectories, providing a more intelligent trajectory generation solution. High-precision and real-time trajectory information generation effectively improves vehicle safety and efficiency, supports the construction and optimization of intelligent transportation systems, and enhances overall traffic management.

Claims

1. A vehicle trajectory data adaptive processing method integrating vehicle and cloud computing, characterized in that: Includes the following steps: S1, Data Acquisition; On the vehicle side, data acquisition equipment is used to collect vehicle trajectory data. The collected vehicle trajectory data includes the vehicle's spatial coordinates and timestamps, and the collected vehicle trajectory data is transmitted to the vehicle's computing device through the vehicle's internal data bus or wireless communication. S2, Data Preprocessing; The vehicle's computing equipment performs preliminary processing on the vehicle trajectory data, including the following sub-steps: S201, Redundancy merging; Set a time window for redundant data and merge vehicle trajectory points with a time difference less than that time window. S202, Null value filling; using adaptive window technology to dynamically adjust the sampling window according to the density of vehicle trajectory data; S203, Data Upload; Upload the pre-processed data to the cloud via wireless network; S3, Data Association; includes the following sub-steps: S301, Spatial Association; Using geospatial indexes to achieve spatial data retrieval and management, spatially association is performed between vehicle trajectory points after redundancy merging and null value filling and road network data, traffic element information and regional POI information. S302, data slice; Geospatial data is sliced ​​according to a preset grid size to achieve gridded management of spatial data; S303, Outlier Correction: An adaptive outlier filtering algorithm based on a combination of sample entropy and local outlier factor is used to identify outliers in vehicle trajectories. S304, Vehicle trajectory information generation; merge the corrected vehicle trajectory points with the normal point set to generate vehicle trajectory information; When executing S202, encrypted data points are used for special operating conditions. Based on the concept of entropy, the complexity of vehicle trajectory data is calculated. For two-dimensional trajectory coordinates, vector difference is used to perform sample entropy analysis to describe the complexity of instantaneous trajectory changes, capture nonlinear features in the signal, and by judging the degree of change in sample entropy within a certain range, a threshold is set to adaptively and dynamically adjust the window size and fill null values. The process also includes the following steps: First, set the initial size to L. S The sampling window is slid towards the vehicle trajectory sequence and the sample entropy of the vehicle trajectory vector in each segment window is calculated. Secondly, a threshold is set, and the sample entropy of adjacent segments is subtracted pairwise; segments whose subtraction results exceed the threshold are considered segments with drastic changes in vehicle trajectory curvature. Next, obtain the maximum length L that satisfies the threshold judgment condition. max And calculate 1-L respectively max The mean sample entropy corresponding to the vehicle trajectory; Next, sort the sample entropy mean from smallest to largest, draw a scatter plot, and connect the minimum and maximum values ​​to form a straight line; Finally, calculate the distance from each point to the line in turn. l Maximum distance l max The corresponding x-coordinate is the optimal window length L; For a given vehicle trajectory sequence B = {b1, b2, K, b...} of length N, ... n The formula for calculating sample entropy is as follows: ； In the formula, integers 0 ≤ m ≤N represents the length of the sequence to be compared, a real number. r > 0 indicates a tolerance level for accepting a match. Indicates that two sequences match m The empirical probability of a point is obtained by the following formula: ; In the formula, C i m ( r The following formula is used to obtain: ; In the formula, C i m ( r ) represents any vector x m ( j )exist x m ( i )of r The probability within the range, NumT ( x m ( i ), x m ( j ))express x m ( i )and x m ( j The distance relationship is less than r Number of times, x m (i) and x m (j) is reconstructed from B. m Dimensional vector.

2. The vehicle trajectory data adaptive processing method integrating vehicle and cloud as described in claim 1, characterized in that: The data acquisition equipment includes video surveillance, inductive loops, radio frequency identification (RFID), automotive electronic identification, and on-board diagnostic systems and GPS devices installed in intelligent vehicles.

3. The vehicle trajectory data adaptive processing method integrating vehicle and cloud as described in claim 1, characterized in that: When executing S201, the vehicle trajectory data is set as: {( x i , y i , t i )}, x i and y i Represents the spatial coordinates of the vehicle. t i Representing timestamps, a time window Δt is set for redundant data. Vehicle trajectory data points with a time difference less than Δt are merged, and the position of the merged vehicle trajectory point is the weighted average of all vehicle trajectory data points within the time window Δt. ； In the formula, w i It's weight. w i The value can be set to 1 or according to the reliability of the data points.

4. The vehicle trajectory data adaptive processing method integrating vehicle and cloud as described in claim 1, characterized in that: Use linear interpolation to fill in the missing values. Let two known vehicle trajectory points be... and The missing point in the middle is ,but: , ; After obtaining L, the time window slides by a certain step size, and adaptive processing of the next segment of the vehicle trajectory sequence begins.

5. The vehicle trajectory data adaptive processing method integrating vehicle and cloud as described in claim 1, characterized in that: When executing S302, the entire geographic space is divided into n× An n-grid, the size of each grid cell Vehicle trajectory points The network index to which it belongs is: , ; When executing S303, the outlier degree of the data point in the local region is adaptively calculated based on the density of neighboring points around the data point, including the following steps: First, find the k nearest neighbors of the target point d and calculate their reachability distance RD. The reachability distance RD is defined as follows: , In the formula, k dis Let d be the distance from point d to its k-th nearest neighbor. d dis ( d, d j () is the distance from point d to d j The distance, j∈{1, 2, …, k }; Secondly, the vehicle trajectory is traversed and the reciprocal of the mean (LRD) of the RD value of each point is calculated and stored in set R1. Next, the vehicle trajectory is traversed again, and the LOF of each point is calculated using R1 and stored in set L1, resulting in: , Finally, based on the threshold c The L1 values ​​are evaluated and filtered to obtain the final target outlier set.

6. The vehicle trajectory data adaptive processing method integrating vehicle and cloud as described in claim 1, characterized in that: Let the set of vehicle trajectory points be {( x i , y i , t i The outlier is} Then the final target outlier can be obtained by correcting the two outliers before and after, as shown in the following formula: , 。 7. The vehicle trajectory data adaptive processing method integrating vehicle and cloud as described in any one of claims 1 to 6, characterized in that: The cloud analyzes and models historical trajectory data to generate predicted trajectories.

Images