Automobile data recorder GPS trajectory tracking and dynamic path analysis method

Abstract

The invention discloses a GPS trajectory tracking and dynamic path analysis method for an automobile data recorder. The method comprises the following steps: S1, collecting and preprocessing vehicle position and timestamp data; s2, constructing an initial trajectory point sequence, and generating a trajectory data set; s3, constructing a multi-section trajectory model, and realizing path dynamic tracking and spatial reconstruction; s4, mapping a current vehicle driving path, and identifying an offset point and an abnormal behavior according to a track fragment matching strategy; s5, extracting vehicle track offset characteristics, and calculating an angle offset accumulated value to generate a deviation score in combination with a control point change trend; s6, triggering a rule engine to adjust a path based on the score, and sending an optimization result to the driving assistance terminal; s7, establishing a multi-segment track model and a video timestamp index, and realizing track and video synchronous playback; and S8, uploading an analysis result to a terminal and a background to realize path monitoring and behavior feedback. The method improves the track precision, the recognition accuracy and the path optimization efficiency, and is suitable for intelligent driving dynamic decision making.

Description

Technical Field

[0001] The present invention relates to the technical field of satellite positioning and trajectory calculation, and particularly to a method for GPS trajectory tracking and dynamic path analysis of a driving recorder. Background Art

[0002] Under the background of the rapid development of the fields of intelligent transportation, Internet of Vehicles, and driving safety, driving recorders have become one of the standard configurations of modern vehicles and are widely used in scenarios such as accident liability determination, driving behavior retrospection, and road traffic supervision. The main functions of traditional driving recorders are concentrated on video recording, which can continuously capture and store the driving pictures in front of the vehicle. Some high-end devices also integrate functions such as sound, collision detection, and gravity sensors. However, with the improvement of traffic safety requirements, relying solely on video data has become difficult to meet the needs of multi-dimensional data fusion such as precise vehicle positioning, behavior trajectory analysis, and intelligent auxiliary decision-making. Therefore, more and more in-vehicle devices have started to integrate GPS positioning modules to achieve real-time recording and trajectory storage of the vehicle's driving position, thereby improving information integrity and event restoration capabilities.

[0003] Although the existing driving recorders combined with GPS functions have basic trajectory recording capabilities, their processing methods are still relatively primitive. Most products only provide simple superposition displays of GPS data. For example, the corresponding longitude and latitude information is displayed on the video playback interface, or static historical driving trajectory lines are drawn on the map. This method is insufficient when facing advanced requirements such as complex road environments, long-term driving behavior analysis, and trajectory anomaly identification. Its main defect lies in the rough trajectory processing method, lacking the ability to clean, model, track, and analyze trajectory data. Especially when facing trajectory data in large ranges, multiple time periods, and multiple scenarios, it is difficult to achieve effective data extraction and pattern analysis.

[0004] In addition, most of the current trajectory analysis algorithms in the market are applied to mobile Internet positioning services and logistics vehicle scheduling systems. Their algorithm designs are often rule-driven or offline batch processing, lacking a response mechanism for the trajectory evolution characteristics in dynamic environments and being difficult to adapt to in-vehicle application scenarios with strong real-time requirements. For example, for the behaviors of vehicles frequently changing lanes, making U-turns, and detouring in complex road conditions, conventional trajectory line drawing methods are prone to errors such as jitter, jump, and drift, and even unable to accurately restore the driving path. These problems lead to a decrease in the accuracy and reliability of trajectory data, thereby reducing the application value of the entire system.

[0005] On the other hand, in terms of anomaly detection, existing solutions usually rely on single-dimensional judgments such as path similarity and speed mutation, lacking the comprehensive analysis ability that combines geometric shape features and temporal features, and are unable to accurately identify subtle deviations or irregular behaviors in the trajectory. For example, when a vehicle deviates from the normal path for a short distance due to avoiding an obstacle, the system may misjudge it as normal driving and fail to trigger a warning; conversely, a reasonable deviation in a complex environment may be misreported as an anomaly, thus affecting the driving experience and subsequent decision-making support.

[0006] In addition, there are obvious defects in the linkage mechanism between the trajectory analysis result and the driving video. Currently, most driving recorders do not establish an indexing relationship between the trajectory and the video, and are unable to retrieve the corresponding video segment from a specific trajectory point, resulting in the user having to manually match the timeline when tracing a specific driving behavior, which is cumbersome and inefficient. This data fragmentation problem makes it difficult for trajectory behavior analysis to form a closed loop with visual evidence, and is not conducive to accident restoration and liability determination.

[0007] Regarding the upload and subsequent application of trajectory data, existing technologies often only support simple local storage, lacking a unified format and upload mechanism, and it is difficult to build a driving behavior database with trajectory data as the core. The lack of a systematic trajectory behavior feedback mechanism will limit the in-depth application of data in fields such as insurance pricing, fleet management, and traffic optimization.

[0008] Therefore, how to provide a method for a driving recorder to track GPS trajectories and analyze dynamic paths to achieve accurate reconstruction of vehicle driving trajectories, dynamic behavior monitoring, abnormal path recognition, and efficient association of trajectories and videos has become an urgent problem for those skilled in the art. Summary of the Invention

[0009] An object of the present invention is to propose a method for a driving recorder to track GPS trajectories and analyze dynamic paths. The present invention makes full use of technical means such as trajectory data cleaning, piecewise interpolation modeling, Bezier curve fitting, trajectory segment matching, and deviation scoring, and details the whole process of realizing real-time trajectory tracking, dynamic path analysis, and abnormal behavior recognition during driving, with the advantages of high path restoration accuracy, accurate deviation recognition, strong data linkage, and applicability to multi-scenario driving analysis.

[0010] According to an embodiment of the present invention, a method for a driving recorder to track GPS trajectories and analyze dynamic paths includes the following steps:

[0011] S1. Collect the GPS position information and timestamp data of the vehicle, and perform preprocessing;

[0012] S2. Concatenate the GPS location information in chronological order of timestamps to construct an initial sequence of trajectory points, and perform noise filtering and smoothing processing on the sequence of trajectory points to generate a cleaned trajectory dataset;

[0013] S3. Based on the trajectory dataset, construct a multi-segment trajectory model based on piecewise linear interpolation and Bezier curve fitting to dynamically track the vehicle driving path and reconstruct the spatial trajectory;

[0014] S4. Map the current vehicle driving path into the multi-segment trajectory model, and identify path deviation points and local abnormal driving behaviors through a trajectory segment matching strategy;

[0015] S5. According to the trajectory segment matching results, extract the vehicle trajectory deviation features, and combine with the change trend of the control points of each segment of the curve in the multi-segment trajectory model, and use the angle deviation cumulative quantization method to score the deviation of the current trajectory;

[0016] S6. According to the trajectory deviation score results, adopt a path adjustment strategy based on a rule engine to optimize the vehicle driving path, and send the optimization results to the driving assistance terminal;

[0017] S7. Establish an index relationship between the multi-segment trajectory model and the timestamps of the driving recorder video to achieve synchronous playback of the trajectory and video and behavior traceability;

[0018] S8. Upload the path analysis results and matching data to the in-vehicle terminal and the background data center to complete dynamic path monitoring and trajectory behavior feedback.

[0019] Optionally, the preprocessing includes time series alignment, outlier removal, and interpolation and completion of missing points.

[0020] Optionally, the S2 specifically includes:

[0021] S21. Sort the collected GPS location information in chronological order of timestamps to generate an ordered sequence of trajectory points;

[0022] S22. Perform speed mutation detection on each trajectory point in the sequence of trajectory points, and remove the trajectory points with abnormal speed changes;

[0023] S23. Use the sliding window average algorithm to smooth the sequence of trajectory points to remove trajectory jitter;

[0024] S24. Perform interpolation operations on the discontinuous sequence of trajectory points to complete the missing positions and generate a cleaned trajectory dataset.

[0025] Optionally, the speed mutation detection refers to calculating the speed between two adjacent trajectory points, identifying speed mutation points by comparing the difference between adjacent speeds with a set speed change threshold, and when the speed difference is greater than the speed change threshold, determining the corresponding trajectory point as an abnormal point and removing it from the trajectory point sequence.

[0026] Optionally, the specific steps of S3 are as follows:

[0027] S31. Denote the cleaned trajectory data set as the trajectory point sequence T = {P1, P2,..., P n}, where P i = (x i , y i , t i ) represents the i-th trajectory point, x i represents the longitude of the i-th trajectory point, y i represents the latitude of the i-th trajectory point, n represents the total number of trajectory points, and t i represents the timestamp of the i-th trajectory point, arranged in ascending order of time;

[0028] S32. Segment the trajectory point sequence T, and generate a set of trajectory sub-segments S = {T1, T2,..., T m} at a fixed time interval Δt, where represents the j-th trajectory sub-segment, and m represents the total number of sub-segments;

[0029] S33. For each trajectory sub-segment T j , use the piecewise linear interpolation method for trajectory fitting. Suppose the trajectory sub-segment contains c points, then the fitting function of the j-th segment is expressed as:

[0030]

[0031] Among them, L j (t) represents the linear interpolation function of the j-th sub-segment at the τ moment, τ is the interpolation time point, satisfying t i ≤ τ ≤ t i+1 , x i and x i+1 respectively represent the longitude values of the i-th and the (i + 1)-th trajectory points, t i and t i+1 respectively represent the timestamps of the i-th and the (i + 1)-th trajectory points, and c represents the number of trajectory points in the current sub-segment;

[0032] S34. For each trajectory sub-segment T j , apply the second-order Bezier curve for smoothing fitting on the basis of interpolation, and construct the second-order Bezier curve function:

[0033] B j (t) = (1 - t)2 ·P a +2(1 - t)t·P b +t 2 ·P c ;

[0034] Among them, B j (t) represents the Bessel curve function at the normalized time t, P a =(x a , y a ) represents the starting point coordinates of the sub - segment, P b =(x b , y b ) represents the control point coordinates, which are used to determine the degree of curve bending, P c =(x c , y c ) represents the ending point coordinates of the sub - segment, x a , x b , x c are the longitude coordinates of the starting point, control point, and ending point respectively, y a , y b , y c are the latitude coordinates of the starting point, control point, and ending point respectively, t represents the normalized time variable, satisfying 0 ≤ t ≤ 1;

[0035] S35. Weightedly fuse each Bessel curve trajectory point B j (t) with the linear interpolation trajectory position L j (t) to form a smooth trajectory segment M j (t), and splice all M j (t) to generate a complete multi - segment trajectory model M(T)={M1(t), M2(t),..., M m (t)}, which serves as the path basis for subsequent dynamic tracking and deviation analysis.

[0036] Optionally, the smooth trajectory segment M j (t) is formed by weightedly fusing the Bessel curve function B j (t) and the linear interpolation function L j (t): For each trajectory sub - segment T j , weightedly fuse the corresponding linear interpolation function L j (t) and the Bessel curve function B j (t) according to the weight factor α ∈ [0, 1] to construct a smooth trajectory segment M j (t)=α·B j (t)+(1 - α)·L j (t)+(1 - α)·L j(t), where α is adaptively adjusted according to the trajectory point direction change rate and speed fluctuation amplitude to enhance the curve's response ability to the trajectory change trend and improve the overall trajectory smoothness.

[0037] Optionally, the S4 specifically includes:

[0038] S41. Obtain the actual trajectory data of the current vehicle within the time period [t s , t e , and form a trajectory sequence T′ = {P′1, P′2,..., P′ q}, where P′ i = (x′ i , y′ i , t′ i ), x′ i represents the longitude of the i-th trajectory point, y′ i represents the latitude of the i-th trajectory point, q represents the total number of current actual trajectory points, and t′ i represents the timestamp of the i-th trajectory point;

[0039] S42. Select the sub-trajectory segments with time matching from the multi-segment trajectory model M(T) = {M1(t), M2(t),..., M m (t)} and perform corresponding relationship matching on the trajectory points according to the timestamps to form trajectory point pairs (P′ , M i , M j (t′ i ));

[0040] S43. Based on the trajectory point pairs (P′ i , M j (t′ i ), construct the trajectory offset function of the current trajectory segment for path offset point identification:

[0041]

[0042] where D(t) represents the trajectory offset function, q represents the number of trajectory points of the current trajectory segment, represents the coordinates of the fitted trajectory point at the moment of t′ i , represents the coordinates of the fitted trajectory point at the moment of t′ i-1 , α′ i , respectively represent the direction angles of the actual trajectory point and the fitted trajectory point, represents the direction angle offset, v′ i , respectively represent the speeds of the actual trajectory points and the fitted trajectory points, λ1 and λ2 are the weight of the direction angle difference and the weight of the speed difference, arctan(·) represents the arctangent function, x′ i-1 represents the longitude of the (i - 1)-th trajectory point, y′ i-1 represents the latitude of the (i - 1)-th trajectory point;

[0043] S44. Compare the trajectory deviation function D(t) with the preset threshold θ d If D(t) > θ d , mark the current trajectory segment as a path deviation segment, and extract the starting and ending points of the trajectory to form a path deviation point set;

[0044] S45. Based on the rate of change of the direction angle |α′ i - α′ i-1 |, the rate of change of speed |v′ i - v′ i-1 | and the continuous change time in each trajectory segment, identify local abnormal driving behaviors such as sharp turn features, frequent speed change features, and abnormal driving pause features, and form an abnormal behavior marking sequence.

[0045] Optionally, the specific content of S5 includes:

[0046] S51. Based on the path deviation point set and the abnormal behavior marking sequence, perform correlation matching on the two sets to construct a deviation feature set E = {e1, e2,..., e r}), where r represents the total number of deviation behaviors, and each deviation behavior e k = (t k , d k , a k , v k ), respectively representing the timestamp t k of the k-th deviation behavior, the spatial deviation d k , the direction deviation angle a k , and the speed deviation value v k ;

[0047] S52. Construct a trajectory deviation comprehensive scoring function S d for measuring the deviation degree of the overall trajectory:

[0048]

[0049] Among them, r represents the total number of deviation behaviors, D = {d1, d2,..., d r} represents the set composed of all spatial deviations, max(D) represents the maximum spatial deviation in the set D, V = {v1, v2,..., v rLet \(V\) denote the set of all speed offset values, \(\max(V)\) denote the maximum speed offset value in set \(V\), and \(w_1, w_2, w_3\) be the weight coefficients corresponding to spatial offset, direction offset, and speed offset respectively, satisfying \(w_1 + w_2 + w_3=1\);

[0050] S53. Based on the control point sequences of each B-spline curve in the multi-segment trajectory model \(M(T)\) Construct a control point perturbation function \(S\) c , which is used to evaluate the morphological change degree between the current trajectory and the multi-segment trajectory model benchmark:

[0051]

[0052] where \(m\) is the number of B-spline curve segments in the multi-segment trajectory model, respectively represent the starting point coordinates, control point coordinates, and ending point coordinates of the \(j\)-th B-spline curve, represents the theoretical control point coordinates of the B-spline curve, generated based on the standard path modeling, represents the B-spline three-point included angle, represents the square of the Euclidean distance between the actual position and the theoretical position of the control point, represents the square of the Euclidean distance between the starting point and the ending point, and \(\gamma_1,\gamma_2\) are the weighted factors of control point perturbation and angle deviation;

[0053] S54. Normalize and fuse the scoring function \(S\) d and the perturbation function \(S\) c to calculate the final trajectory deviation score \(S\) f \(=\eta\cdot S\) d +(1 - \(\eta\))\(\cdot S\) c , where \(\eta\) is the scoring fusion factor, representing the fusion weight of behavior and trajectory structure scoring, and is used as the input basis for subsequent path adjustment strategies.

[0054] Optionally, the specific content of S51 includes:

[0055] S511. Represent the path offset point set as a set \(D =\{(t\) k ,d\) k )\}, where \(t\) k represents the timestamp of the \(k\)-th offset point, and \(d\) k represents the spatial offset of the \(k\)-th offset point;

[0056] S512. Represent the abnormal behavior marker sequence as a set \(A=\{(t\) k ,a\) k ,v\) k )\}, where \(a\) k represents the direction offset angle at timestamp \(t\) k , and \(v\) k represents the timestamp \(t\)k Speed offset value;

[0057] S513. Through the timestamp t k Associate and match the two sets to construct an offset feature set E as the input basis for trajectory deviation scoring.

[0058] Optionally, the S6 specifically includes:

[0059] S61. Obtain the trajectory deviation score S f , and compare the score value with the path adjustment threshold θ s . If S f is greater than θ s , trigger the path optimization process;

[0060] S62. Extract the current vehicle position L c =(x c , y c ), the current speed v c , the current direction angle α c and the road information set R={r1, r2,..., r z} within the future estimated time window Δt, where each r z represents a feasible path segment;

[0061] S63. Based on the offset feature set E={e1, e2,..., e r}, the deviation score S f and the current position L c , construct a path adjustment candidate set P={p1, p2,..., p u}, where each p u represents an alternative path that meets the current driving conditions;

[0062] S64. Calculate the path length, section congestion weight, historical offset frequency and current behavior pattern matching degree for each candidate path p u respectively, and combine the path smoothness coefficient μ, path safety coefficient ν, and path preference factor ρ in the driving objective function to construct a path score sequence;

[0063] S65. Select the path p * with the optimal path score in P as the adjustment path and convert it into segmented navigation instructions, where the instructions include direction, distance and operation time;

[0064] S66. Send the path instructions to the driving assistance terminal and give real-time prompts to the driver through the interface presentation method, including voice broadcast and path guidance.

[0065] The beneficial effects of the present invention are:

[0066] First, the present invention preprocesses GPS trajectory data, including time series alignment, outlier removal, missing value filling, and trajectory smoothing, effectively improving the continuity and accuracy of trajectory data, solving problems such as trajectory point jumps and data drift in the prior art, and laying a stable data foundation for subsequent modeling and analysis. The method of constructing a multi-segment trajectory model by combining piecewise linear interpolation and Bezier curve fitting not only improves the spatial fitting accuracy of trajectory expression but also has the ability of dynamic update, can track the driving path in real time during vehicle driving, and meets the actual requirements of high-real-time trajectory modeling.

[0067] Secondly, through the trajectory segment matching strategy, the present invention makes a fine-grained comparison of the spatial and direction changes between the current trajectory and the modeled trajectory, and can identify the path deviation and abnormal driving behavior of the vehicle in real time. Combined with the angle deviation cumulative quantization method, an accurate trajectory deviation scoring mechanism is established, effectively improving the accuracy and robustness of deviation detection, and solving the problems of false judgment and missed judgment caused by only relying on speed or simple similarity calculation in the prior art. At the same time, the path adjustment strategy using a rule engine enables the system to intelligently recommend driving paths according to the degree of deviation, enhancing the practicality and decision-making assistance ability of trajectory analysis.

[0068] Finally, the present invention establishes an index mapping relationship between the trajectory model and the video time stamp of the driving recorder, realizes the linked display of the trajectory and video data, enables users to accurately locate to the corresponding video frame when tracing back a certain abnormal trajectory behavior, and significantly improves the event tracing efficiency. At the same time, it supports the upload of trajectory behavior data to in-vehicle terminals and the background data center, realizes dynamic path monitoring, driving behavior feedback, and data archiving, provides structured and high-value data support for building an intelligent traffic analysis platform, and has broad application prospects and engineering promotion value. Description of the Drawings

[0069] The drawings are used to provide a further understanding of the present invention, and constitute a part of the specification, and are used to explain the present invention together with the embodiments of the present invention, and do not constitute a limitation to the present invention. In the drawings:

[0070] Figure 1 is a flowchart of a method for GPS trajectory tracking and dynamic path analysis of a driving recorder proposed by the present invention;

[0071] Figure 2 is a flowchart of constructing a multi-segment trajectory model of a method for GPS trajectory tracking and dynamic path analysis of a driving recorder proposed by the present invention;

[0072] Figure 3 is a flowchart of trajectory deviation scoring of a method for GPS trajectory tracking and dynamic path analysis of a driving recorder proposed by the present invention. Detailed Embodiments

[0073] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are all simplified schematic diagrams, only showing the basic structure of the present invention in a schematic manner, so they only show the components related to the present invention.

[0074] Reference Figures 1 - 3 , a method for GPS trajectory tracking and dynamic path analysis of a driving recorder, comprising the following steps:

[0075] S1. Collect the GPS position information and timestamp data of the vehicle, and perform preprocessing;

[0076] S2. Stitch the GPS position information in the order of timestamps to construct an initial trajectory point sequence, and perform noise filtering and smoothing processing on the trajectory point sequence to generate a cleaned trajectory data set;

[0077] S3. Based on the trajectory data set, construct a multi-segment trajectory model based on piecewise linear interpolation and Bezier curve fitting to perform dynamic tracking and spatial trajectory reconstruction of the vehicle driving path;

[0078] S4. Map the current vehicle driving path into the multi-segment trajectory model, and identify path deviation points and local abnormal driving behaviors through a trajectory segment matching strategy;

[0079] S5. According to the trajectory segment matching result, extract the vehicle trajectory deviation characteristics, and combine the change trend of the control points of each segment of the curve in the multi-segment trajectory model, and use the angle deviation cumulative quantization method to score the deviation of the current trajectory;

[0080] S6. According to the trajectory deviation scoring result, adopt a path adjustment strategy based on a rule engine to optimize the vehicle driving path, and send the optimization result to the driving assistance terminal;

[0081] S7. Establish an index relationship between the multi-segment trajectory model and the driving recorder video timestamp to realize synchronous playback and behavior traceability of the trajectory and the video;

[0082] S8. Upload the path analysis result and matching data to the in-vehicle terminal and the background data center to complete dynamic path monitoring and trajectory behavior feedback.

[0083] By deeply integrating the GPS trajectory data with the driving recorder system, the present invention designs a complete processing flow from data collection, trajectory modeling, deviation recognition, dynamic scoring to path adjustment, realizing real-time tracking of the vehicle trajectory and dynamic path analysis. This method can effectively restore the actual driving path, identify deviation behaviors in real time, and generate personalized path optimization suggestions, improving the practical value and technical level of the driving recorder in intelligent driving assistance.

[0084] In this embodiment, the preprocessing includes time series alignment, outlier removal, and interpolation for missing points.

[0085] By introducing time series alignment, outlier removal, and interpolation for missing points in the trajectory preprocessing stage, the present invention significantly improves the continuity, stability, and accuracy of trajectory data, provides a clean and reliable data basis for subsequent trajectory modeling and analysis, and effectively reduces the modeling deviation caused by errors in the original GPS data.

[0086] In this embodiment, S2 specifically includes:

[0087] S21. Sort the collected GPS position information in timestamp order to generate an ordered trajectory point sequence;

[0088] S22. Detect speed mutations for each trajectory point in the trajectory point sequence, and remove the trajectory points with abnormal speed changes;

[0089] S23. Smooth the trajectory point sequence using the sliding window average algorithm to remove trajectory jitters;

[0090] S24. Perform interpolation operations on the discontinuous trajectory point sequence to fill in the missing positions and generate a cleaned trajectory data set.

[0091] The present invention refines the trajectory stitching and data cleaning processes. Through steps such as sorting, speed mutation detection, sliding window smoothing, and interpolation, it ensures the consistency and stability of the trajectory point sequence in both time and space dimensions, further enhancing the accuracy of the trajectory model and the robustness of the system.

[0092] In this embodiment, the speed mutation detection refers to calculating the speed between two adjacent trajectory points, identifying speed mutation points by comparing the difference between adjacent speeds with a set speed change threshold. When the speed difference is greater than the speed change threshold, the corresponding trajectory point is determined as an abnormal point and removed from the trajectory point sequence.

[0093] Through the speed mutation detection mechanism, the present invention can effectively remove abnormal trajectory points caused by GPS drift, signal occlusion, or device anomalies, thus avoiding interference from incorrect data in the overall trajectory analysis process and enhancing the system's adaptability to sudden speed anomalies.

[0094] In this embodiment, S3 specifically includes:

[0095] S31. Denote the cleaned trajectory data set as the trajectory point sequence T = {P1, P2,..., P n}, where P i = (x i , y i , ti ) represents the i-th trajectory point, x i represents the longitude of the i-th trajectory point, y i represents the latitude of the i-th trajectory point, n represents the total number of trajectory points, t i represents the timestamp of the i-th trajectory point, arranged in ascending order of time;

[0096] S32. Segment the trajectory point sequence T, and generate a set of trajectory sub-segments S = {T1, T2,..., T m} at a fixed time interval Δt, where represents the j-th trajectory sub-segment, and m represents the total number of sub-segments;

[0097] S33. For each trajectory sub-segment T j , use the piecewise linear interpolation method for trajectory fitting. Suppose the trajectory sub-segment contains c points, then the fitting function for the j-th segment is expressed as:

[0098]

[0099] where, L j (t) represents the linear interpolation function of the j-th sub-segment at the τ moment, τ is the interpolation time point, satisfying t i ≤ τ ≤ t i+1 , x i and x i+1 respectively represent the longitude values of the i-th and the (i + 1)-th trajectory points, t i and t i+1 respectively represent the timestamps of the i-th and the (i + 1)-th trajectory points, and c represents the number of trajectory points in the current sub-segment;

[0100] S34. For each trajectory sub-segment T j , apply a second-order Bezier curve for smoothing fitting on the basis of interpolation, and construct a second-order Bezier curve function:

[0101] B j (t) = (1 - t) 2 · P a + 2(1 - t)t · P b + t 2 · P c ;

[0102] where, B j (t) represents the Bezier curve function at the normalized time t, P a = (x a , y a ) represents the starting point coordinates of the sub-segment, P b = (x b , y b ) represents the control point coordinates, which are used to determine the degree of curve bending, Pc =(x c , y c ) represents the termination point coordinates of the sub-segment. x a , x b , x c are the longitude coordinates of the starting point, control point, and ending point respectively. y a , y b , y c are the latitude coordinates of the starting point, control point, and ending point respectively. t represents the normalized time variable, satisfying 0 ≤ t ≤ 1;

[0103] S35. Weightedly fuse each Bezier curve trajectory point B j (t) with the linear interpolation trajectory position L j (t) to form a smooth trajectory segment M j (t), and splice all M j (t) to generate a complete multi-segment trajectory model M(T) = {M1(t), M2(t),..., M m (t)}, which serves as the path basis for subsequent dynamic tracking and deviation analysis.

[0104] By introducing a double modeling mechanism of piecewise linear interpolation and Bezier curve fitting, the present invention constructs a multi-segment trajectory model, which can realize the spatial smooth expression and dynamic update of the vehicle trajectory, solves the deficiencies of the traditional trajectory model in terms of continuity, flexibility, and responsiveness, and improves the authenticity of the trajectory expression and the expansion ability of the system.

[0105] In this embodiment, the smooth trajectory segment M j (t) is formed by weightedly fusing the Bezier curve function B j (t) and the linear interpolation function L j (t): For each trajectory sub-segment T j , weightedly fuse the corresponding linear interpolation function L j (t) and the Bezier curve function B j (t) according to the weight factor α ∈ [0, 1] to construct a smooth trajectory segment M j (t) = α · B j (t) + (1 - α) · L j (t), where α is adaptively adjusted according to the trajectory point direction change rate and speed fluctuation amplitude to enhance the response ability of the curve to the trajectory change trend and improve the overall trajectory smoothness. j

[0106] ​By weighted fusion of the linear interpolation function and the Bezier curve function, and introduction of adaptive factors such as speed fluctuation and direction change rate, the present invention can achieve dynamic response to different driving behavior characteristics, effectively improve the adaptability and stability of the trajectory model in the actual driving scenario, and enhance the smoothness and expressiveness of the trajectory reconstruction.

[0107] In this embodiment, step S4 specifically includes:

[0108] S41. Obtain the actual trajectory data of the current vehicle within the time period [t s , t e to form a trajectory sequence T′ = {P′1, P′2,..., P′ q}, where P′ i = (x′ i , y′ i , t′ i ), x′ i represents the longitude of the i-th trajectory point, y′ i represents the latitude of the i-th trajectory point, q represents the total number of current actual trajectory points, and t′ i represents the timestamp of the i-th trajectory point;

[0109] S42. Select a sub-trajectory segment with time matching from the multi-segment trajectory model M(T) = {M1(t), M2(t),..., M m (t)} and perform corresponding relationship matching on the trajectory points according to the timestamps to form trajectory point pairs (P′ , M i , M j (t′ i ));

[0110] S43. Based on the trajectory point pairs (P′ i , M j (t′ i )) of the current trajectory segment, construct a trajectory offset function of the current trajectory segment for path offset point identification:

[0111]

[0112] where D(t) represents the trajectory offset function, q represents the number of trajectory points of the current trajectory segment, represents the coordinates of the fitted trajectory point at the moment of t′ i , represents the coordinates of the fitted trajectory point at the moment of t′ i-1 , α′ i , respectively represent the direction angles of the actual trajectory point and the fitted trajectory point, represents the direction angle offset, v′ i , respectively represent the speeds of the actual trajectory points and the fitted trajectory points, λ1 and λ2 are the weight of the direction angle difference and the weight of the speed difference, arctan(·) represents the arctangent function, and x′ i-1 represents the longitude of the (i - 1)-th trajectory point, and y′ i-1 represents the latitude of the (i - 1)-th trajectory point;

[0113] S44. Compare the trajectory deviation function D(t) with the preset threshold θ d If D(t) > θ d , mark the current trajectory segment as a path deviation segment, and extract the starting and ending points of the trajectory to form a path deviation point set;

[0114] S45. Based on the rate of change of the direction angle |α′ i - α′ i-1 |, the rate of change of the speed |v′ i - v′ i-1 | and the continuous change time in each trajectory segment, identify local abnormal driving behaviors such as sharp turn features, frequent speed change features, and abnormal driving pause features, and form an abnormal behavior marking sequence.

[0115] Through the time alignment and spatial comparison of the current trajectory and the modeled trajectory segment, the present invention establishes a trajectory deviation function model, and combines the speed and direction angle changes to analyze the path deviation behavior, which can accurately identify minute trajectory deviations and abnormal driving behaviors, providing a quantitative basis for subsequent deviation evaluation and behavior marking.

[0116] In this embodiment, the specific steps of S5 are as follows:

[0117] S51. Based on the path deviation point set and the abnormal behavior marking sequence, perform correlation matching on the two sets to construct a deviation feature set E = {e1, e2,..., e r}, where r represents the total number of deviation behaviors, and each deviation behavior e k = (t k , d k , a k , v k ), which respectively represent the timestamp t k of the k-th deviation behavior, the spatial deviation d k , the direction deviation angle a k , and the speed deviation value v k ;

[0118] S52. Construct a comprehensive trajectory deviation scoring function S d for measuring the deviation degree of the overall trajectory:

[0119]

[0120] where r represents the total number of offset behaviors, D = {d1, d2,..., d r} represents the set composed of all spatial offsets, max(D) represents the maximum spatial offset in the set D, V = {v1, v2,..., v r} represents the set composed of all speed offset values, max(V) represents the maximum speed offset value in the set V, w1, w2, and w3 are the weight coefficients corresponding to spatial offset, direction offset, and speed offset respectively, and satisfy w1 + w2 + w3 = 1;

[0121] S53. Based on the control point sequences of each B-spline curve in the multi-segment trajectory model M(T) Construct a control point perturbation function S c , which is used to evaluate the morphological change degree between the current trajectory and the multi-segment trajectory model benchmark:

[0122]

[0123] where m is the number of B-spline curve segments in the multi-segment trajectory model, respectively represent the starting point coordinates, control point coordinates, and ending point coordinates of the j-th B-spline curve, represents the theoretical control point coordinates of the B-spline curve, generated based on the standard path modeling, represents the included angle of three points of the B-spline curve, represents the square of the Euclidean distance between the actual position and the theoretical position of the control point, represents the square of the Euclidean distance between the starting point and the ending point, and γ1, γ2 are the weighting factors of control point perturbation and angle deviation;

[0124] S54. Normalize and fuse the scoring function S d and the perturbation function S c to calculate the final trajectory deviation score S f = η·S d +(1 - η)·S c , where η is the scoring fusion factor, representing the fusion weight of behavior and trajectory structure scoring, and is used as the input basis for subsequent path adjustment strategies.

[0125] The present invention forms a multi-dimensional scoring system for trajectory spatial offset, direction angle change, and model structure perturbation by constructing an offset feature set, a trajectory deviation scoring function, and a control point perturbation function, effectively enhancing the comprehensiveness and accuracy of trajectory anomaly assessment, and providing a scientific basis for path decision-making and safety monitoring.

[0126] In this embodiment, the S51 specifically includes:

[0127] S511. Represent the path offset point set as the set D = {(tk , d k )}, where t k represents the timestamp of the k-th offset point, and d k represents the spatial offset of the k-th offset point;

[0128] S512. Represent the abnormal behavior marker sequence as a set A = {(t k , a k , v k )}, where a k represents the direction offset angle of the timestamp t k , and v k represents the speed offset value of the timestamp t k ;

[0129] S513. Through the timestamp t k , perform associated matching on the two sets to construct an offset feature set E as the input basis for trajectory deviation scoring.

[0130] By precisely defining the set structures of path offset points and abnormal behavior points and performing efficient matching based on timestamps, the present invention can fuse and analyze spatio-temporal offset features and behavior features, realize the structured expression of complex driving behaviors, and improve the interpretability and guidance of offset detection results in subsequent path analysis.

[0131] In this embodiment, the S6 specifically includes:

[0132] S61. Obtain the trajectory deviation score S f , and compare the score value with the path adjustment threshold θ s . If S f is greater than θ s , trigger the path optimization process;

[0133] S62. Extract the current vehicle position L c = (x c , y c ), the current speed v c , the current direction angle α c , and the road information set R = {r1, r2,..., r z} within the future estimated time window Δt, where each r z represents a feasible path segment;

[0134] S63. Based on the offset feature set E = {e1, e2,..., e r}, the deviation score S f , and the current position L c , construct a path adjustment candidate set P = {p1, p2,..., p u}, and each pu Indicates an alternative route that meets the current driving conditions;

[0135] S64. For each candidate route p u Calculate the route length, road segment congestion weight, historical deviation frequency, and the matching degree with the current behavior pattern respectively, and combine the route smoothness coefficient μ, route safety coefficient ν, and route preference factor ρ in the driving objective function to construct a route score sequence;

[0136] S65. Select the route p with the optimal route score * ∈P as the adjusted route, and convert it into segmented navigation instructions, where the instructions include direction, distance, and operation time;

[0137] S66. Send the route instructions to the driving assistance terminal and give real-time prompts to the driver through the interface presentation, including voice broadcast and route guidance.

[0138] Based on obtaining the trajectory score, the path adjustment strategy of the present invention comprehensively considers the real-time position, speed, direction, and road topology information, constructs a path candidate set and a scoring mechanism, and selects the optimal path through the constraint of the driving objective function, which can realize dynamic and personalized path recommendation; at the same time, combined with the linkage of navigation instructions and the driving assistance terminal, it improves the intelligent guidance ability of the system and the user interaction experience.

[0139] Embodiment 1:

[0140] To verify the feasibility of the present invention in implementation, the present invention is applied to the operation vehicle monitoring and management of an intelligent bus dispatching system in a certain city. In this city, there are problems such as serious road congestion during morning and evening rush hours, frequent deviation of bus vehicles from the preset routes, and difficulty in tracking and supervising driving behaviors. Although the conventional driving recorder equipment can provide video evidence, it lacks real-time path monitoring and behavior analysis capabilities and is difficult to meet the actual requirements of intelligent transportation dispatching for trajectory accuracy and dynamic analysis.

[0141] Therefore, we select 8 main lines of the BRT rapid bus system in this city as the pilot, and each line is equipped with 3 bus vehicles installed with the system of the present invention. The test period lasts for 15 days in total. The vehicles are respectively deployed with multifunctional driving recorders embedded with GPS modules and video recording units, and integrate the trajectory tracking and dynamic analysis method proposed by the present invention to realize local processing and remote synchronization of data. From 6:00 in the morning to 21:00 in the evening every day is the continuous operation period. During this period, the system automatically executes complete processes such as GPS data acquisition, trajectory preprocessing, multi-segment trajectory modeling, deviation identification, and path adjustment.

[0142] During actual operation, the system of the present invention first collects and cleans the GPS data during vehicle driving, solving the drift phenomenon caused by high-rise building occlusion, weather interference, etc. in conventional GPS positioning. For example, at 8:43 am on March 12, the system detected that the vehicle numbered BRT-03 had a trajectory jump when passing through the A3 elevated ramp, and the original longitude and latitude deviation was as high as 26 meters. The system successfully corrected this section of the trajectory into a continuous curve and constructed a smooth trajectory segment through time series alignment and trajectory interpolation algorithms.

[0143] In terms of path deviation recognition, the system can accurately capture deviation behaviors based on the matching of multi-segment trajectory models and real-time trajectories. For example, at 17:28 pm on March 15, the vehicle BRT-05 had a temporary detour when passing through West Z1 Road. The system detected that the deviation distance was 14.6 meters, the deviation duration was 73 seconds, and the trajectory deviation score reached 0.82, which was higher than the set warning threshold of 0.75. This triggered the path adjustment module and pushed an alternative path suggestion to the driving terminal to help the driver quickly return to the main line.

[0144] In terms of driving behavior analysis, the system identifies abnormal behaviors through features such as the speed change rate within the trajectory segment, the direction angle change frequency, and the stop duration. Taking March 17 as an example, the system detected a total of 21 sharp turn behaviors, 36 frequent acceleration and deceleration behaviors, and 8 abnormal parking behaviors. Most of them occurred near the vehicle entering a nearby station or a temporary dispatching point, indicating that the driving behavior is highly correlated with the station density and dispatching tasks.

[0145] Through the comprehensive statistics of 15 days of data, the system recorded a total of 143 trajectory deviation events, among which there were only 9 false alarms, and the deviation recognition accuracy rate reached 93.7%; the automatic association accuracy between the trajectory and video data reached 100%, and all deviation behaviors could be quickly traced back to the corresponding video clips in the interface; the average response time for all vehicle path adjustments did not exceed 3 seconds, and the average retrieval delay of the video and trajectory playback interface was less than 0.9 seconds, significantly better than the average time-consuming of 1.5 minutes for traditional manual retrieval.

[0146] The present invention not only effectively solves the problems of difficult real-time tracking of bus trajectories, many misjudgments in deviation detection, and slow response in path adjustment, but also improves the dispatching efficiency and operation safety through system-level integration, providing a highly reliable, high-precision, and expandable trajectory behavior analysis platform for the intelligent bus system. Compared with traditional trajectory recording schemes, the present invention can significantly improve the trajectory accuracy, abnormal recognition ability, and system linkage, and has stronger practicality and promotion value.

[0147] The following table is the result record of the recognition of the running trajectory deviation and path optimization feedback of some bus vehicles during the test period. The table data comes from the logs automatically generated by the actual deployment system and has high authenticity and representativeness:

[0148] Table 1 Statistical Table of Identification of Trajectory Deviation and Feedback of Route Adjustment for Smart Bus

[0149]

[0150] As can be seen from the above embodiments, the present invention shows extremely high trajectory fitting accuracy, abnormal identification ability and route optimization effect in the urban bus trajectory monitoring scenario. It not only solves the problems of rough trajectory processing and lack of dynamic judgment mechanism in the traditional driving recorder system, but also realizes the automation and visualization of trajectory-video linkage backtracking, significantly enhancing the application value of trajectory data in the intelligent transportation system.

[0151] The above is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution and inventive concept of the present invention, makes equivalent substitutions or changes, and should be covered by the protection scope of the present invention.

Claims

1. A method for GPS trajectory tracking and dynamic path analysis of a driving recorder, characterized in that, It includes the following steps: S1. Collect the GPS position information and timestamp data of the vehicle, and perform preprocessing; S2. Stitch the GPS position information according to the timestamp order to construct an initial trajectory point sequence, and perform noise filtering and smoothing processing on the trajectory point sequence to generate a cleaned trajectory dataset; S3. Based on the trajectory dataset, construct a multi-segment trajectory model based on piecewise linear interpolation and Bezier curve fitting to dynamically track the vehicle driving path and reconstruct the spatial trajectory; S4. Map the current vehicle driving path into the multi-segment trajectory model, and identify path deviation points and local abnormal driving behaviors through a trajectory segment matching strategy; S5. According to the trajectory segment matching result, extract the vehicle trajectory deviation feature, combine the change trend of the control points of each segment curve in the multi-segment trajectory model, and use the angle deviation cumulative quantization method to score the deviation of the current trajectory; S6. According to the trajectory deviation scoring result, adopt a path adjustment strategy based on a rule engine to optimize the vehicle driving path, and send the optimization result to the driving assistance terminal; S7. Establish an index relationship between the multi-segment trajectory model and the video timestamp of the driving recorder to realize synchronous playback of the trajectory and video and behavior traceability; S8. Upload the path analysis result and matching data to the in-vehicle terminal and the background data center to complete dynamic path monitoring and trajectory behavior feedback.

2. A method for GPS trajectory tracking and dynamic path analysis of a driving recorder according to claim 1, characterized in that, The preprocessing includes time series alignment, outlier removal, and interpolation and completion of missing points.

3. A method for GPS trajectory tracking and dynamic path analysis of a driving recorder according to claim 1, characterized in that, The specific content of S2 includes: S21. Sort the collected GPS position information according to the timestamp order to generate an ordered trajectory point sequence; S22. Perform speed mutation detection on each trajectory point in the trajectory point sequence, and remove the trajectory points with abnormal speed changes; S23. Use the sliding window average algorithm to smooth the trajectory point sequence to remove trajectory jitter; S24. Perform interpolation operations on the discontinuous trajectory point sequence to complete the missing positions and generate a cleaned trajectory dataset.

4. A method for GPS trajectory tracking and dynamic path analysis of a driving recorder according to claim 3, characterized in that, The speed mutation detection refers to calculating the speed between two adjacent trajectory points, identifying the speed mutation points by comparing the difference between adjacent speeds and the set speed change threshold. When the speed difference is greater than the speed change threshold, the corresponding trajectory point is determined as an abnormal point and removed from the trajectory point sequence.

5. A method for GPS trajectory tracking and dynamic path analysis of a driving recorder according to claim 1, characterized in that, The specific content of S3 includes: S31. Denote the cleaned trajectory dataset as the trajectory point sequence T = {P1, P2,..., P n}, where P i = (x i , y i , t i ) represents the i-th trajectory point, x i represents the longitude of the i-th trajectory point, y i represents the latitude of the i-th trajectory point, n represents the total number of trajectory points, and t i represents the timestamp of the i-th trajectory point, arranged in ascending order of time; S32. Segment the trajectory point sequence T and generate a set of trajectory sub - segments S = {T1, T2,..., T m} at a fixed time interval Δt, where represents the j - th trajectory sub - segment and m represents the total number of sub - segments; S33. For each trajectory sub-segment T j , the piecewise linear interpolation method is used for trajectory fitting. Suppose the trajectory sub-segment contains c points, then the fitting function of the j-th segment is expressed as: Among them, L j (t) represents the linear interpolation function of the j-th sub-segment at the τ moment, where τ is the interpolation time point, satisfying t i ≤τ≤t i+1 , x i and x i+1 represent the longitude values of the i-th and (i + 1)-th trajectory points respectively, t i and t i+1 represent the timestamps of the i-th and (i + 1)-th trajectory points respectively, and c represents the number of trajectory points in the current sub-segment; S34. For each trajectory sub-segment T j , on the basis of interpolation, apply a second-order Bezier curve for smooth fitting to construct a second-order Bezier curve function: B j (t) = (1 - t) 2 ·P a + 2(1 - t)t·P b + t 2 ·P c ; Among them, B j (t) represents the Bessel curve function at the normalized time t, and P a =(x a , y a ) represents the starting point coordinates of the sub-segment, and P b =(x b , y b ) represents the control point coordinates, which are used to determine the degree of curve bending. P c =(x c , y c ) represents the ending point coordinates of the sub-segment. x a , x b , x c are the longitude coordinates of the starting point, the control point, and the ending point respectively. y a , y b , y c are the latitude coordinates of the starting point, the control point, and the ending point respectively. t represents the normalized time variable, satisfying 0 ≤ t ≤ 1; S35. Weightedly fuse each Bessel curve trajectory point B j (t) with the linear interpolation trajectory position L j (t). Form a smooth trajectory segment M j (t), and splice all M j (t) to generate a complete multi-segment trajectory model M(T) = {M1(t), M2(t),..., M m (t)}, which serves as the path basis for subsequent dynamic tracking and deviation analysis.

6. A method for GPS trajectory tracking and dynamic path analysis of a driving recorder according to claim 5, characterized in that The smooth trajectory segment M j (t) is the Bezier curve function B j (t) and the linear interpolation function L j (t) Weighted fusion formation: For each trajectory sub-segment T j , T j The corresponding linear interpolation function L j (t) and Bezier curve function B j (t) Perform weighted fusion according to the weight factor α∈[0,1] to construct a smooth trajectory segment M j (t) = α·B j (t)+(1-α)·L j (t), where α is adaptively adjusted according to the direction change rate and speed fluctuation amplitude of the trajectory point to enhance the responsiveness of the curve to the trajectory change trend and improve the overall trajectory smoothness.

7. A method for GPS trajectory tracking and dynamic path analysis of a driving recorder according to claim 1, characterized in that, The specific content of S4 includes: S41. Obtain the actual trajectory data of the current vehicle within the time period [t s , t e , and form a trajectory sequence T′ = {P′1, P′2,..., P′ q}, where P′ i = (x′ i , y′ i , t′ i ), x′ i represents the longitude of the i-th trajectory point, y′ i represents the latitude of the i-th trajectory point, q represents the total number of current actual trajectory points, and t′ i represents the timestamp of the i-th trajectory point; S42. Select sub-trajectory segments with time matching from the multi-segment trajectory model M(T) = {M1(t), M2(t),..., M m (t)} And match the corresponding relationships of the trajectory points according to the timestamps to form trajectory point pairs (P′ i , M j (t′ i )); S43. Based on the trajectory point pair (P′ i , M j (t′ i ))), construct the trajectory offset function of the current trajectory segment for path offset point recognition: Among them, D(t) represents the trajectory offset function, q represents the number of trajectory points of the current trajectory segment, represents the coordinates of the fitted trajectory point at time t′ i moment, represents the coordinates of the fitted trajectory point at time t′ i-1 moment, respectively represent the direction angles of the actual trajectory point and the fitted trajectory point, represents the direction angle offset, respectively represent the velocities of the actual trajectory point and the fitted trajectory point, λ1, λ2 are the direction angle difference weight and the velocity difference weight, arctan(·) represents the arctangent function, x′ i-1 represents the longitude of the (i - 1)-th trajectory point, y′ i-1 represents the latitude of the (i - 1)-th trajectory point; S44. Compare the trajectory offset function D(t) with a preset threshold θ d If D(t) > θ d , mark the current trajectory segment as a path offset segment, and extract the starting and ending points of the trajectory to form a path offset point set; S45. Based on the rate of change of the direction angle |α′ i -α′ i-1 |, the rate of change of speed |v′ i -v′ i-1 |, and the continuous change time, identify the local abnormal driving behaviors of sharp turn features, frequent speed change features, and abnormal driving pause features, and form an abnormal behavior marker sequence.

8. A method for GPS trajectory tracking and dynamic path analysis of a driving recorder according to claim 1, characterized in that, The specific content of S5 includes: S51. Based on the path offset point set and the abnormal behavior marking sequence, associate and match the two sets to construct an offset feature set E = {e1, e2,..., e r}, where r represents the total number of offset behaviors, and each offset behavior e k = (t k , d k , a k , v k ), representing the timestamp t k , spatial offset d k , direction offset angle a k , and speed offset value v k of the k-th offset behavior respectively; S52. Construct a comprehensive trajectory deviation scoring function S d , which is used to measure the deviation degree of the overall trajectory: Among them, r represents the total number of offset behaviors, D = {d1, d2,..., d r} represents the set composed of all spatial offsets, max(D) represents the maximum spatial offset in the set D, V = {v1, v2,..., v r} represents the set composed of all speed offset values, max(V) represents the maximum speed offset value in the set V, and w1, w2, w3 are the weight coefficients corresponding to spatial offset, direction offset, and speed offset respectively, satisfying w1 + w2 + w3 = 1; S53. Based on the control point sequences of each Bezier curve in the multi-segment trajectory model M(T) Construct a control point perturbation function S c , which is used to evaluate the degree of morphological change between the current trajectory and the multi-segment trajectory model benchmark: where m is the number of Bezier curve segments in the multi-segment trajectory model, respectively represent the starting point coordinates, control point coordinates, and ending point coordinates of the j-th Bezier curve, represents the theoretical control point coordinates of the Bezier curve, generated based on the standard path modeling, represents the included angle of three points of the Bezier curve, represents the square of the Euclidean distance between the actual position and the theoretical position of the control point, represents the square of the Euclidean distance between the starting point and the ending point, and γ1, γ2 are the weighting factors of the control point perturbation and the angle deviation; S54. Normalize and fuse the scoring function S d and the perturbation function S c to calculate the final trajectory deviation score S f = η · S d + (1 - η) · S c , where η is the scoring fusion factor, representing the fusion weight of the behavior and trajectory structure scores, and is used as the input basis for subsequent path adjustment strategies.

9. A method for GPS trajectory tracking and dynamic path analysis of a driving recorder according to claim 8, characterized in that, The specific content of S51 includes: S511. Represent the path offset point set as a set D = {(t k , d k )}, where t k represents the timestamp of the k-th offset point, and d k represents the spatial offset of the k-th offset point; S512. Represent the abnormal behavior marking sequence as a set \(A =\{(t k ,a k ,v k )\}, where \(a k represents the direction offset angle of timestamp \(t k , and \(v k represents the speed offset value of timestamp \(t k ;\) S513. By means of timestamp t k Perform correlation matching on two sets to construct an offset feature set E as the input basis for trajectory deviation scoring.

10. A method for GPS trajectory tracking and dynamic path analysis of a driving recorder according to claim 1, characterized in that, The specific content of S6 includes: S61. Obtain the trajectory deviation score S f , and compare the score value with the path adjustment threshold θ s . If S f is greater than θ s , trigger the path optimization process; S62. Extract the current vehicle position L c =(x c , y c ), the current speed v c , the current direction angle α c and the road information set R = {r1, r2,..., r z} within the future estimated time window Δt, where each r z represents a feasible path segment; S63. Based on the offset feature set E = {e1, e2,..., e r}, the deviation score S f and the current position L c , construct a path adjustment candidate set P = {p1, p2,..., p u}, where each p u represents an alternative path that satisfies the current driving conditions; S64. For each candidate path p u Calculate the path length, road segment congestion weight, historical deviation frequency, and the matching degree with the current behavior pattern respectively, and combine the path smoothness coefficient μ, path safety coefficient ν, and path preference factor ρ in the driving objective function to construct a path score sequence; S65. Select the path p with the optimal path score * ∈P as the adjustment path and convert it into segmented navigation instructions, where the instructions include direction, distance, and operation time; S66. Send the path instruction to the driving assistance terminal, and give real-time prompts to the driver through the interface presentation method, including voice broadcast and path guidance.

Images