Method for identifying suspicious behaviors based on circling trajectory

By extracting and analyzing the action trajectory characteristics from the flow trajectory data, identifying hovering behaviors and combining community environment characteristics, the problem of difficult identification of abnormal behavior trajectories in community correction is solved, efficient identification and monitoring of suspicious behaviors is achieved, and work efficiency is improved .

CN119066302AInactive Publication Date: 2024-12-03GUIYANG INFORMATION TECH RES INST
View PDF 5 Cites 0 Cited by

Patent Information

Application Number
CN202411565727.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-11-05
Publication Date
2024-12-03
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

In community correction work, it is difficult for the existing technology to effectively identify and monitor abnormal behavior trajectories, resulting in staff spending a lot of time to screen out a very small number of abnormal behavior data from massive data.

Method used

By extracting action trajectories and features from the flow trajectory data, calculating symbol horns and velocity-related indicators, identifying hovering behavior, and combining community environmental characteristics to calculate the degree of suspicious behavior to identify suspicious behavior. Specific steps include obtaining streaming trajectory data, extracting action trajectories and features, identifying hovering behavior, calculating the degree of suspicious behavior, and identifying suspicious behavior based on suspiciousness weights.

Benefits of technology

It improves the accuracy and efficiency of identifying suspicious behaviors, saves staff's time and energy, and improves the efficiency of community correction work.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119066302A_ABST
    Figure CN119066302A_ABST
Patent Text Reader

Abstract

The invention discloses a method for identifying suspicious behaviors based on a spiral trajectory, which comprises the following steps of: acquiring streaming trajectory data, and extracting an action trajectory and action characteristics; wherein the ordered action position of the moving object forms an action track S (C1... Cn); the action characteristics comprise time, speed and distance of a moving object; a symbol angle is calculated from the action trajectory, and a circling behavior is recognized and comprises a circling trajectory and a speed related to the circling trajectory; calculating a behavior suspicious degree according to the circling behavior; and identifying the suspicious behavior according to the behavior suspicious degree and the suspicious weight. According to the technical scheme, the effectiveness and accuracy of the recognition result of the suspicious behavior can be improved, the time and energy of workers can be effectively saved, and the working efficiency is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of location information big data applications, and in particular to a method for identifying suspicious behaviors based on circling trajectories. Background Art

[0002] In recent years, mobile phone positioning technology has been widely used in community correction work to monitor and record the movement trajectory of correction subjects. In the application, communication base stations can be used to determine the location of community correction personnel. Communication base stations have a certain coverage range between each other. When the correction personnel's mobile phones switch within the signal range of the base station, their locations can be determined, thereby obtaining the behavior trajectory of the target personnel to achieve real-time monitoring of the target personnel. During real-time monitoring, staff are required to monitor dynamic movement trajectories and identify abnormal movement trajectories. However, usually, most of the movement trajectories of target personnel are regular and normal behaviors, and the trajectory data corresponding to abnormal actions is a very small part of all movement data. With the rapid growth of monitoring information data, it takes a lot of time and effort for staff to effectively identify abnormal behavior data from massive data. Therefore, it is necessary to use technical means to combine the characteristics of community environment and suspicious behavior to conduct preliminary identification and early warning of movement trajectories.

[0003] According to the analysis of community events that have occurred in recent years: before an incident occurs, suspicious persons usually stop and wander around certain areas that need to be focused on (i.e., stations, schools, and kindergartens) at a very slow speed to observe the environment. It can be seen that the trajectory data of abnormal behavior has certain characteristics. In recent years, many research results have been produced in the field of abnormal trajectory detection: for example, outlier trajectory detection methods for frequent paths based on time constraints and spatial factors, but although such methods improve the accuracy of abnormal trajectory detection, they do not combine the characteristics of abnormal behavior trajectory data, and are not suitable for monitoring application scenarios of community corrections. Therefore, a technical solution is needed that can identify abnormal trajectories in combination with the characteristics of abnormal behavior trajectories and apply them to community corrections. Summary of the invention

[0004] To achieve the above objectives, the present application provides a method for identifying suspicious behavior based on circling trajectories, comprising the following steps: Obtain streaming trajectory data, and extract action trajectories and action features from the streaming trajectory data; wherein the streaming trajectory data is an ordered data set composed of the chronological order of the mobile object during the action process; the streaming trajectory data includes time T i and time T i Corresponding location information point C i (x i ,y i ); The ordered action positions of the moving objects constitute the action trajectory S (C1...Cn ); the action characteristics include the time, speed and distance of the moving object; Calculate the symbol angle from the motion trajectory to identify the circling behavior, the circling behavior includes the circling trajectory and the speed related to the circling trajectory; wherein the motion trajectory of a moving object has multiple ordered motion positions with the same position information, the motion trajectory between two ordered motion positions with the same position information is a single-endpoint trajectory, and the circular single-endpoint trajectory is a circling trajectory; Calculate the suspiciousness of behavior based on circling behavior; Identify suspicious behaviors based on the suspicious degree and suspicious weight of the behaviors.

[0005] The extraction of the action trajectory from the streaming trajectory data includes the following steps: Define neighborhood radius , neighborhood radius Used to determine whether two location information points are adjacent; Define the minimum number of location information points MinPts within the domain radius; Extract i core points from the streaming trajectory data, where the number of neighbors is greater than or equal to MinPts as the core point; Determine the adjacent core points connected by density to form the action trajectory S(C1...C n ),.

[0006] Furthermore, identifying the hovering behavior from the action trajectory includes the following steps: Extract three consecutive position information points C from a single endpoint trajectory j-1 , C j and C j+1 Composition of C j The first vector and the second vector , wherein the number of position information points of the single-endpoint trajectory is n; According to the C j The first vector and the second vector Get C j Symbol Angle ; Traverse all information location points in the single-end point trajectory and obtain the symbol angles of all location information points; if , the single-endpoint trajectory is a spiral trajectory.

[0007] Among them, the symbol angle The calculation method is: , in: , .

[0008] Further, calculating the suspicious degree of the behavior according to the hovering behavior includes the following steps: Specify a motion trajectory, obtain the circling trajectory of the motion trajectory and the speed associated with the circling trajectory, the speed associated with the circling trajectory includes the average speed of the moving target in the motion trajectory , the standard deviation is expressed as , the average speed of the circling trajectory , calculate the suspicious speed value SSV; Get the sign angle of the hovering trajectory and calculate the hovering suspicious value CSV; Furthermore, the suspicious degree SV of the action trajectory is calculated as follows: .

[0009] The calculation method of the speed suspicious value SSV is: .

[0010] The calculation method of the hovering suspicious value CSV is: .

[0011] Furthermore, according to the suspicious degree and suspicious weight of the behavior, identifying suspicious behavior includes the following steps: Acquire an observation target of the hovering trajectory, acquire environmental features of the observation target, and determine a suspicious weight; updating the suspicious degree of the behavior according to the suspicious weight; A preset standard threshold is extracted, and action trajectories with a suspicious degree greater than the standard threshold are extracted, and the behavior of the mobile target corresponding to the action trajectory is identified as suspicious behavior.

[0012] The calculation method of updating the suspiciousness of the behavior according to the suspiciousness weight is as follows: , where w is the weighting coefficient.

[0013] Furthermore, before identifying the circling behavior from the action trajectory, determining whether the action trajectory needs to identify the circling behavior includes the following steps: Extract any two position information points A(x A ,y A )、B(x B ,y B ); Determine any two position information points A(x A ,y A )、B(x B ,y B) between the vector distance ratio VD; When the vector distance ratio VD is less than the specified threshold, the action trajectory needs to identify the circling behavior.

[0014] Among them, any two location information points A(x A ,y A )、B(x B ,y B ) is calculated as follows: ; in, For any two location information points A(x A ,y A )、B(x B ,y B ), through A(x A ,y A )、B(x B ,y B ) between the action trajectories S(C i ) in the location information point C i ( ) is calculated using the following method: , where n is the number of location information points; dist(A,B)is the position information point A(x A ,y A )、B(x B ,y B )) is calculated as: .

[0015] According to the present invention, the effectiveness and accuracy of the identification results of suspicious behaviors can be improved, and the quantified suspicious degree of behavior can be applied to the work of community corrections, which can effectively save the time and energy of staff and improve work efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 is a step diagram of a method for identifying suspicious behavior based on circling trajectories provided according to an embodiment of the present invention; Figure 2 is a schematic diagram of a motion trajectory data structure provided according to an embodiment of the present invention; Figure 3 is a schematic diagram of circling behavior provided according to an embodiment of the present invention. DETAILED DESCRIPTION

[0017] The present invention proposes a method for identifying suspicious trajectories from assembly-line trajectory data. First, a clustering method is used to obtain relevant features of the action trajectory from the trajectory data, including the characteristics of the behavior trajectory and the characteristics of the behavior. Then, combined with the community environment, the degree of suspicious behavior of the action corresponding to the trajectory is calculated. The degree of suspicion is a quantitative value and can be applied to community correction work to provide early warning, etc.

[0018] The specific implementation of the present invention is described in detail below with reference to the accompanying drawings.

[0019] Figure 1 A method step diagram for identifying suspicious behavior based on circling trajectories is provided, as shown in the figure, including the following steps: Step S100: Obtain streaming trajectory data; The streaming trajectory data is an ordered data set composed of the chronological order of the moving object during its action. The streaming trajectory data includes the time T i and time T i Corresponding location information point C i (x i ,y i ), where i is a sequence number, (x i ,y i ) is the GPS coordinate point of the location information point, time T i and time T i Corresponding location information point C i (x i ,y i ) constitute an orderly action position.

[0020] The streaming trajectory data involved in the present invention is obtained through a mobile terminal and an integrated data system that have legal authorization to collect data.

[0021] In this step, according to the regularity of occurrence time, such as the continuity of occurrence time, data within the valid time period is further extracted from the streaming trajectory data and structured. The structure of the extracted data is as follows: Figure 2 As shown: Extract action trajectories and action features from streaming trajectory data. The action trajectory includes time T i and the corresponding location information point C i (x i ,y i ); Position information point C in continuous time i The set of ordered action positions of mobile objects constitutes the action trajectory S (C1...C n ), where n is the number of ordered action positions; action features include the time, speed, and distance of the moving object. Further, in subsequent calculations, according to the time T iand distance, extracting features such as average speed; In the present invention, the DBSCAN algorithm is used to extract the action trajectory from the streaming trajectory data. DBSCAN (Density-Based Spatial Clustering of Applications with Noise) is an algorithm based on high-density continuous regions, which can identify clusters of any shape and automatically determine the number of clusters; DBSCAN can quickly calculate cluster points, concentrate the data of irregularly distributed location information points, select continuous location information points, and generate action trajectories.

[0022] Extracting action trajectories includes the following steps: 1) Define the neighborhood radius : Among them, the neighborhood radius Used to determine whether two location information points are adjacent through coordinate points: when the distance between the two location information points is less than the neighborhood radius , then these two location information points are neighbors; 2) Define the minimum number of location information points MinPts within the domain radius; that is, define the neighborhood radius Minimum number of neighbors within 3) Extract i core points from the streaming trajectory data, where the location information points whose number of neighbors is greater than or equal to MinPts are core points, and the core points are location information points whose number of neighbors is not less than MinPts; 4) Determine whether adjacent core points are densely connected. The densely connected core points constitute the action trajectory S (C i ): In this step, we first determine whether two adjacent core points are density-reachable. On the basis of density reachability, the adjacent core points connected by density constitute the action trajectory S (C1...C n ).

[0023] The process of extracting action trajectories can be implemented through the following pseudo code: / / Input trajectory dataset Input: a trajectory Points=[ ... ]; / / m is the number of location information points of streaming trajectory data / / Output action trajectory Output: a set of candidates trajectory points –>CPoints begin / / Call algorithm AdaptiveClusterMethod AdaptiveClusterMethod (Points, n) / / Define an empty list All_clusterlist←empty list; All_noiselist←empty list selectiveID←1; MinPtsList←empty list / / Initialization parameters max = sizeof(Points) MinPtsList←1 to max / / Traverse the trajectory dataset for each(MinPtsMinPtsList) do SubsetgList←empty list ε←Get ε(Points, MinPts) / / Call DBSCAN to obtain cluster point set SubsetList←DBSCAN(ε, MinPts) Assign MinPts to( SubsetSubsetList) add(CSubsetList) into All_clusterlist add(NSubsetList) into All_noiselis return GetClusteringResult (Points, All_clusterlist, All_noiselist, n, selectiveID) / / Output cluster point set as location information points in the action trajectory, where n is the number of location information points in the action trajectory End Through the above method, the action trajectory S(C1...C n ).

[0024] For some suspicious behaviors, the person may try to hover or linger around the point of interest, and try multiple routes, thus generating a route with circling characteristics. Generally speaking, such a route with circling characteristics will form a circular trajectory. Therefore, in the present invention, after obtaining the action trajectory, the action trajectory is analyzed to see whether there is circling behavior, that is, the circling behavior of the action trajectory is identified.

[0025] Step S110: Identify circling behavior from the action trajectory: The circling behavior includes the circling trajectory and speed information related to the circling trajectory, such as the average speed of the circling trajectory. ; Among them, the action trajectory of a moving object has multiple ordered action positions with the same position information (that is, the trajectories intersect), the action trajectory between two ordered action positions with the same position information with the smallest trajectory distance is a single-endpoint trajectory, and the circular single-endpoint trajectory is a spiral trajectory; The average speed of the moving object in the spiral trajectory is the average speed of the spiral trajectory .

[0026] Unlike normal trajectories, suspicious trajectories will generate redundant travel distances in the trajectories from two points because the person trying to hover around the point of interest will try to hover around the point of interest. Therefore, before executing this step, the vector distance ratio is used to determine whether the action trajectory has redundant distances to determine whether the circling behavior needs to be identified in the action trajectory, including the following steps: Extract any two position information points A(x A ,y A )、Point B(x B ,y B ); Determine the two arbitrary position information points A(x A ,y A )、Point B(x B ,y B ) between the vector distance ratio VD, The calculation method is: ; in, For any two location information points A(x A ,y A )、B(x B ,y B ), through A(x A ,y A )、B(x B ,y B ) between the action trajectories S(C i ) in the location information point C i ( ) is calculated using the following method: , where n is the number of location information points; dist(A,B)is any two location information points A(x A ,y A )、B(x B ,y B )) is calculated as: .

[0027] The range of VD is 0 to 1. When VD is close to 0, it means that the action trajectory has a redundant distance from point A to point B. When VD is close to 1, it means that the action trajectory goes directly from point A to point B. In this step, a threshold is defined in advance, for example, the threshold is specified as 0.2. When the vector distance ratio VD is less than the specified threshold, the action trajectory needs to be identified as a circling behavior.

[0028] If the action trajectory needs to identify circling behavior, perform the following steps: 1) Extract three consecutive location information points C from a single endpoint trajectory j-1 , C j and C j+1 Composition of C j The first vector and the second vector , where j is the sequence number and the number of location information points of a single endpoint trajectory is n; 2) According to C j The first vector and the second vector Get C j Symbol Angle ; Symbol Angle The calculation method is: , in: Used to calculate angles, sign() is the sign function: if )>0,sign( ) is equal to 1; if )<0,sign( ) is equal to -1, if ) =0,sign( ) is equal to 0; in: , .

[0029] 3) Traverse all information location points in the single-endpoint trajectory and obtain the symbol angles of all location information points; if , the single-endpoint trajectory is a spiral trajectory.

[0030] That is: when the sum of the absolute values ​​of the symbol angles in the trajectory is greater than or equal to When , the trajectory is determined to be a detour, such as Figure 3 As shown, the trajectory contains 8 points and 5 vector angles. and is positive, to is negative, the geometric relationship is as follows: ;in, and in the opposite direction, Greater than or equal to , the starting point of the trajectory loop is set to C1. In addition, is positive, so , therefore, from Start to The ending partial trajectory is defined as a spiral trajectory.

[0031] The calculation process can be implemented by the following pseudo code: Algorithm Name: CircleMotifsMethod Input: candicate trajectory CPoints [ , … ] Output: a set of circling motifs of CPoints(CM),the sum of angle( Begin CM =[], start =, end = null, ; for j from 1 to n-1 do { Algorithm Name: CircleMotifsMethod Input: candicate trajectory CPoints [ , … ] Output: a set of circling motifs of CPoints(CM),the sum of angle( Begin CM =[], start =, end = null, ; for j from 1 to n-1 do { if 0<| |<3 / 2 and sign( ) – sign( )>0 then start=i,end= null, = ; else if| |>3 / 2 then end =j } } return [CM, ] End In one action trajectory, there may be multiple spiral trajectories. When there are multiple spiral trajectories, the sum of the absolute values ​​of the symbol angles is greater than or equal to .

[0032] Step S120: Calculating the suspiciousness of the action trajectory according to the circling behavior; After obtaining the circling behavior in the action trajectory, the suspicious degree of the behavior can be calculated based on the characteristics of the action trajectory, including the following steps: Specify a motion trajectory, obtain the circling trajectory of the motion trajectory and the speed related to the circling trajectory, and obtain the average speed of the moving target in the motion trajectory , Standard Deviation , the average speed of the spiral trajectory , on this basis, calculate the speed suspicious value SSV; The calculation method is: ; Get the sign angle of the circling trajectory and calculate the circling suspicious value CSV. The calculation method is: ; The suspicious degree SV of the action trajectory is calculated as follows: .

[0033] The suspicious behavior program SV generated at this time is calculated and generated by the generation characteristics of the action trajectory and has not yet been combined with the community environment.

[0034] Step S130: Identify suspicious behaviors according to the suspicious degree and suspicious weight of the behaviors.

[0035] In this step, the observed target of the hovering trajectory is obtained, the environmental characteristics of the observed target are obtained, and the suspicious weight is determined; Before determining the observation target of the circling trajectory, sensitive monitoring points can be determined according to the characteristics of the community, such as places where minors gather (schools, kindergartens), areas with heavy traffic (hospitals, stations), etc. The observation target is determined by the distance between the circling trajectory and the sensitive monitoring point.

[0036] After determining the observation target of the circling trajectory, determine the suspicious weight, including: using w=1 weighting for sensitive monitoring points, while other areas can be weighted with 0 (w=0).

[0037] In this step, the calculation method of updating the suspicious degree of the behavior according to the suspicious weight is: , where w is the weighting coefficient.

[0038] After the suspicious behavior is updated, the pre-set suspicious behavior standard threshold can be extracted, and the action track with a suspicious behavior level greater than the standard threshold can be extracted, and the action track corresponding to the mobile target is identified as suspicious behavior. After the suspicious behavior is identified by the present invention, an early warning can be issued in the application of community correction, which is convenient for staff to follow up.

[0039] In the present invention, a large amount of traffic trajectory data is analyzed, and the idea of ​​clustering is introduced through the DBSCAN algorithm to locate the core point in the area where a large number of location information points exist, and quickly generate the action trajectory; and based on the distance and angle relationship between each location information point in the action trajectory, the circling trajectory with suspicious properties is located. On this basis, combined with the sensitive areas that need to be monitored in the community, weights are introduced to generate a quantified degree of suspicious behavior. The degree of suspicious behavior calculated by the present invention combines the data characteristics and environmental characteristics of community events, so that the effectiveness and accuracy of the identification results are effectively improved. The quantified degree of suspicious behavior can be conveniently applied to the work of community corrections. The staff can monitor the trajectory with a high degree of suspicious behavior without having to identify abnormal behavior data from massive data, which effectively saves the staff time and energy and improves work efficiency.

[0040] The above disclosures are only several specific embodiments of the present invention; however, the present invention is not limited thereto, and any changes that can be conceived by those skilled in the art should fall within the protection scope of the present invention.

Claims

1. A method for identifying suspicious behavior based on circling trajectories, characterized in that: The following steps are involved: Acquire streaming trajectory data, and extract action trajectories and action features from the streaming trajectory data; wherein the streaming trajectory data is an ordered data set composed of the chronological order of the mobile object during the action process; the streaming trajectory data includes time T i and the time T i Corresponding location information point C i (x i ,y i ); The ordered action positions of the moving objects constitute the action trajectory S (C1...C n ); the action characteristics include the time, speed and distance of the moving object; Calculating the symbol angle from the motion trajectory to identify the circling behavior, wherein the circling behavior includes the circling trajectory and the speed related to the circling trajectory; wherein the motion trajectory of a moving object has multiple orderly motion positions with the same position information, the motion trajectory between two orderly motion positions with the same position information is a single-endpoint trajectory, and the circular single-endpoint trajectory is a circling trajectory; Calculating a suspicious behavior degree based on the circling behavior; Suspicious behaviors are identified based on the suspicious degree and suspicious weight of the behaviors.

2. The method for identifying suspicious behavior based on circling trajectory according to claim 1, characterized in that: The step of extracting the action trajectory from the streaming trajectory data comprises the following steps: Define neighborhood radius , the neighborhood radius Used to determine whether two location information points are adjacent; Define the minimum number of location information points MinPts within the domain radius; Extract i core points from the streaming trajectory data, where the number of neighbors is greater than or equal to MinPts as the core point; Determine the adjacent core points connected by density to form the action trajectory S(C1...C n ).

3. The method for identifying suspicious behavior based on circling trajectory according to claim 1, characterized in that: The step of identifying the circling behavior from the action trajectory comprises the following steps: Extract three consecutive position information points C from a single endpoint trajectory j-1 , C j and C j+1 Composition of C j The first vector and the second vector , wherein the number of position information points of the single-endpoint trajectory is n; According to the C j The first vector and the second vector Get C j Symbol Angle ; Traversing all information position points in the single-end point trajectory to obtain the symbol angles of all position information points; if , the single-endpoint trajectory is a spiral trajectory.

4. The method for identifying suspicious behavior based on circling trajectory according to claim 3, characterized in that: The sign angle The calculation method is: , in: , 。 5. The method for identifying suspicious behavior based on circling trajectory according to claim 1, characterized in that: Calculating the suspicious degree according to the hovering behavior comprises the following steps: Specify a motion trajectory, obtain the circling trajectory of the motion trajectory and the speed associated with the circling trajectory, the speed associated with the circling trajectory includes the average speed of the moving target in the motion trajectory , the standard deviation is expressed as , the average speed of the circling trajectory , calculate the suspicious speed value SSV; Obtaining the sign angle of the circling trajectory and calculating a circling suspicious value CSV; The suspicious behavior level SV of the action trajectory is calculated, and the calculation method is: 。 6. The method for identifying suspicious behavior based on circling trajectory according to claim 5, characterized in that: The calculation method of the speed suspicious value SSV is: ; The calculation method of the hovering suspicious value CSV is: 。 7. The method for identifying suspicious behavior based on circling trajectory according to claim 1, characterized in that: According to the suspicious degree and suspicious weight of the behavior, identifying the suspicious behavior includes the following steps: Acquire an observation target of the hovering trajectory, acquire environmental features of the observation target, and determine a suspicious weight; updating the suspicious degree of the behavior according to the suspicious weight; A preset standard threshold is extracted, and action trajectories with a suspicious degree greater than the standard threshold are extracted, and the behavior of the mobile target corresponding to the action trajectory is identified as suspicious behavior.

8. The method for identifying suspicious behavior based on circling trajectories according to claim 7, characterized in that: The calculation method of updating the suspicious degree of the behavior according to the suspicious weight is: , where w is the weighting coefficient.

9. The method for identifying suspicious behavior based on circling trajectory according to claim 1, characterized in that: Before identifying the circling behavior from the motion trajectory, determining whether the motion trajectory needs to identify the circling behavior includes the following steps: Extract any two position information points A(x A ,y A )、B(x B ,y B ); Determine the two arbitrary position information points A(x A ,y A )、B(x B ,y B ) between the vector distance ratio VD; When the vector distance ratio VD is less than a specified threshold, the action trajectory needs to identify a circling behavior.

10. The method for identifying suspicious behavior based on circling trajectory according to claim 9, characterized in that: Any two location information points A(x A ,y A )、B(x B ,y B ) is calculated as follows: ; in, For any two location information points A(x A ,y A )、B(x B ,y B ), through A(x A ,y A )、B(x B ,y B ) between the action trajectories S(C i ) in the location information point C i ( ) is calculated using the following method: , where n is the number of location information points; dist(A,B)is the position information point A(x A ,y A )、B(x B ,y B )) is calculated as: .

Citation Information

Patent Citations

  • Method for track monitoring in positioning system

    CN103901451A

  • Abnormal behavior analysis method based on target trajectory

    CN113011367A

  • Ship abnormal trajectory detection method and device, computer equipment and storage medium

    CN115510946A

  • Multi-modal trajectory generation method for vertical take-off and landing unmanned aerial vehicle group

    CN117170230A

  • Track-based sensor parameter calibration method and device, and autonomous vehicle

    CN117647782A