Pedestrian and non-motor vehicle track segmentation method and system based on change point detection

By using a change point detection-based method, the feature distance of trajectory points is calculated and trajectory data is segmented, which solves the robustness and accuracy problems of pedestrian and non-motorized vehicle trajectory segmentation in the existing technology, and realizes more accurate analysis of pedestrian and non-motorized vehicle motion status.

CN119027448AActive Publication Date: 2024-11-26SUN YAT SEN UNIV
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
CN202410996568.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-24
Publication Date
2024-11-26
Estimated Expiration
2044-07-24

AI Technical Summary

Technical Problem

Existing technologies are not robust enough for pedestrian and non-motorized vehicle trajectory segmentation, have low segmentation accuracy, and are greatly affected by parameters. They are particularly difficult to effectively segment the movement status of pedestrians and non-motorized vehicles in traffic-intensive urban intersection scenarios.

Method used

By acquiring trajectory data and preprocessing it, calculating the feature distance of trajectory points, identifying points of change in movement patterns, dividing the trajectory data into different sub-trajectories based on the points of change, and determining whether the state is stationary or moving, sub-trajectories with the same state are merged to obtain the trajectory segmentation results for pedestrians and non-motorized vehicles.

Benefits of technology

It enhances the robustness and accuracy of trajectory segmentation, avoids the hassle of manually setting segmentation thresholds, enables a more detailed understanding of the motion state of pedestrians and non-motorized vehicles, and improves data quality and analysis efficiency.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119027448A_ABST
    Figure CN119027448A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of trajectory analysis, and provides a pedestrian and non-motor vehicle trajectory segmentation method and system based on change point detection, and the method comprises the steps: obtaining trajectory data comprising a plurality of trajectory points, and carrying out the preprocessing of the trajectory data; calculating the feature distance of each trajectory point in the trajectory data; based on the characteristic distance of the track point, obtaining the track point with a changed moving mode; dividing the trajectory data into different sub-trajectories based on the trajectory points with the changed moving modes, judging the sub-trajectories to be in a moving or static state, and merging the sub-trajectories in the same state to obtain a pedestrian and non-motor vehicle trajectory segmentation result. Compared with the defects of inaccurate division and large calculation amount in the prior art, the invention provides a segmentation method which is more efficient, accurate and wide in applicability.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of trajectory analysis, and in particular to a pedestrian and non-motor vehicle trajectory segmentation method and system based on change point detection. Background Art

[0002] The development of sensing technology in the past decade has led to an explosive growth in mobile object data. Massive data provides a basis for traffic behavior analysis, and it is necessary to analyze these data to obtain more effective conclusions. In the current traffic environment, pedestrians and non-motor vehicles are usually in a disadvantaged position in traffic, especially in densely populated urban intersections. The travel state and speed of pedestrians and non-motor vehicles are changeable, and their behavior directly affects the traffic safety and efficiency of the entire intersection. Therefore, it is necessary to segment the trajectories of pedestrians and non-motor vehicles and grasp their different movement states, which will help to better understand their behavior patterns and provide strong data support for the formulation of traffic management plans.

[0003] At present, there are many methods for trajectory segmentation research. For example, the trajectory sequence that meets the required spatiotemporal conditions is detected by threshold. Common thresholds include spatial distance, time length, speed and angle change thresholds. Although this method is easy to implement and relatively intuitive, it is not robust to noise. Another method is to segment trajectories based on cost functions. This method constructs the most uniform trajectory segments by minimizing a specific cost function. For example, GRASP-UTS based on the minimum description length principle finds a trajectory segmentation method that can maximize the data compression effect. However, this method has high time complexity and is not suitable for large data sets. In addition, the clustering-based segmentation method is mainly used to describe the behavior of individuals staying in a specific area for more than a period of time. This method detects stops by identifying continuous sequences adjacent in space and time, but this method is greatly affected by clustering parameters and is prone to over-segmentation or incomplete segmentation, especially for mixed trajectories of pedestrians and non-motor vehicles. Summary of the invention

[0004] In order to overcome the defects of the above-mentioned prior art, such as lack of robustness, great influence of parameters and low segmentation accuracy, the present invention provides a pedestrian and non-motor vehicle trajectory segmentation method and system based on change point detection.

[0005] In order to achieve the above technical effects, the technical solution of the present invention is as follows:

[0006] The present invention proposes a pedestrian and non-motor vehicle trajectory segmentation method based on change point detection, comprising the following steps:

[0007] Acquire trajectory data including a number of trajectory points and perform preprocessing;

[0008] Calculating a characteristic distance of each trajectory point in the trajectory data;

[0009] Obtaining a trajectory point where the movement mode changes based on the characteristic distance of the trajectory point;

[0010] The trajectory data is divided into different sub-trajectories based on the trajectory points where the movement mode changes, and the sub-trajectories are judged to be in a moving or stationary state, and the sub-trajectories in the same state are merged to obtain a pedestrian and non-motor vehicle trajectory segmentation result.

[0011] The present invention also proposes a pedestrian and non-motor vehicle trajectory segmentation system based on change point detection, the system comprising:

[0012] A trajectory data preprocessing module is used to preprocess the acquired trajectory data;

[0013] A feature distance calculation module is used to calculate the feature distance of each trajectory point in the trajectory data;

[0014] The trajectory segmentation module is used to calculate the trajectory points where the movement mode changes based on the characteristic distance of each trajectory point, and divide the trajectory data into different sub-trajectories on this basis, and judge them as moving or stationary states, merge sub-trajectories in the same state, and obtain the pedestrian and non-motor vehicle trajectory segmentation results.

[0015] The present invention also proposes an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method for evaluating the optimal ratio of wind and solar energy capacity as described in the present invention when executing the computer program.

[0016] Compared with the prior art, the present invention has the following beneficial effects:

[0017] The present invention is based on a sliding window method. After preprocessing the trajectory data, the characteristic distance of each trajectory point is obtained, and whether the movement mode has changed is determined according to the characteristic distance. The trajectory data is divided into different sub-trajectories based on the trajectory points where the movement mode has changed, and then the sub-trajectories are judged to be moving or stationary. The sub-trajectories in the same state are merged to obtain the result of pedestrian and non-motor vehicle trajectory segmentation. The present invention avoids the trouble of manually setting the segmentation threshold in the existing method by calculating the characteristic distance of each trajectory point and analyzing whether the movement state has changed, thereby enhancing the robustness of the method; at the same time, the trajectory points where the movement state has changed are divided into different sub-trajectories, which can ensure the accuracy of the trajectory division. In addition, by accurately distinguishing the states of different sub-trajectories and dividing them into moving states or stationary states, it is helpful to understand the movement state of pedestrians and non-motor vehicles in more detail. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 This is a flow chart of the pedestrian and non-motor vehicle trajectory segmentation method based on change point detection in Example 1.

[0019] Figure 2 This is an architecture diagram of the pedestrian and non-motor vehicle trajectory segmentation system based on change point detection in Example 2. DETAILED DESCRIPTION

[0020] The accompanying drawings are only used for illustrative purposes and are not to be construed as limiting the present invention;

[0021] It is understandable to those skilled in the art that some well-known descriptions may be omitted in the drawings.

[0022] The technical solution of the present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0023] Example 1

[0024] This embodiment proposes a pedestrian and non-motor vehicle trajectory segmentation method based on change point detection, such as Figure 1 , which is a flow chart of the pedestrian and non-motor vehicle trajectory segmentation method based on change point detection in this embodiment.

[0025] The present embodiment proposes a pedestrian and non-motor vehicle trajectory segmentation method based on change point detection, comprising the following steps:

[0026] Acquire trajectory data including a number of trajectory points and perform preprocessing;

[0027] Calculating a characteristic distance of each trajectory point in the trajectory data;

[0028] Obtaining a trajectory point where the movement mode changes based on the characteristic distance of the trajectory point;

[0029] The trajectory data is divided into different sub-trajectories based on the trajectory points where the movement mode changes, and the sub-trajectories are judged to be in a moving or stationary state, and the sub-trajectories in the same state are merged to obtain a pedestrian and non-motor vehicle trajectory segmentation result.

[0030] In this embodiment, by calculating the characteristic distance of the trajectory points in the trajectory data, the trajectory points where the movement mode changes are determined by the characteristic distance, the trajectory data is divided into different sub-trajectories based on the trajectory points where the movement mode changes, and then the sub-trajectory state is determined to be static or moving, and the sub-trajectories with the same state are merged to obtain the division result. Compared with the prior art, this method does not need to specify the segmentation threshold by itself, which enhances the robustness of the method; for the divided sub-trajectories, a simple and efficient method is used to distinguish the static and moving states, which simplifies the calculation while ensuring the accuracy of trajectory division.

[0031] In an optional embodiment, the trajectory point includes longitude, latitude and time; the preprocessing step includes: analyzing the trajectory data and individual trajectory points and removing trajectory points that do not meet preset requirements.

[0032] In this embodiment, the data collected from the trajectory points only need to include the most basic location information and time. After analyzing the trajectory data and individual trajectory points, removing the trajectory points that do not meet the preset requirements can effectively reduce noise and abnormal data, thereby improving the quality of the overall data and the accuracy of the analysis. In addition, reducing abnormal data can also reduce the computational burden in the data processing process, improve the efficiency and response speed of the system, and thus more effectively identify the behavior patterns of pedestrians and non-motor vehicles.

[0033] As an example, a trajectory point that does not meet the preset requirements is a trajectory point that meets any of the following conditions:

[0034] The number of trajectory points is less than 10;

[0035] Track points lacking time or longitude and latitude;

[0036] Noise points whose distance from adjacent trajectory points exceeds 10m.

[0037] In an optional embodiment, the step of calculating the characteristic distance of each trajectory point in the trajectory data includes:

[0038] For any trajectory point, the trajectory data of a certain distance before and after the trajectory point is selected as its past trajectory and future trajectory respectively;

[0039] The feature vectors of the past trajectory and the future trajectory are calculated, and the feature distance of any trajectory point is calculated based on the feature vectors of the past trajectory and the future trajectory.

[0040] Furthermore, the feature vector includes the average speed, average angle, average density and loop coefficient of all trajectory points in the trajectory data; the average density includes the average number of neighbor points of the trajectory point; the loop coefficient is the ratio of the straight-line distance of the trajectory data to the actual movement distance; its expression is as follows:

[0041]

[0042] Among them, f3(s i ) is the average density, where N(p i ) represents the i-th trajectory point p i The number of neighboring points within the radius R; k represents the number of trajectory points contained in the trajectory data; f4(s i ) is the loop coefficient, and eu(·) represents the Euclidean distance calculation function.

[0043] Specifically, the average speed and average angle of all trajectory points in the trajectory data can be calculated according to the following formula:

[0044]

[0045] Among them, f1(s i ) is the average speed, Speed(p i ) represents the velocity of the i-th trajectory point; f2(s i ) is the average angle, Angle(p i ) represents the angle of the i-th trajectory point.

[0046] Furthermore, the eigenvector is expressed by the following formula:

[0047] F(p i-l+1 →p i )=(f1(s i ),f2(s i ),f3(s i ),f4(s i ))

[0048] Among them, F(p i-l+1 →p i ) represents the feature vector of the trajectory data from the i-l+1th trajectory point to the i-th trajectory point;

[0049] As an exemplary illustration, the feature vector of the past feature is represented by the following formula:

[0050]

[0051] Among them, F past (p i ) is p i The feature vector of the point's past features.

[0052] As another exemplary illustration, the feature vector of the future feature is represented by the following formula:

[0053]

[0054] Among them, F future (p i ) is p i The feature vector of the future features of the point.

[0055] In this embodiment, by selecting past and future features of fixed length through trajectory points, the behavior pattern before and after a specific time point can be analyzed in detail to identify the behavior transition point; by selecting a variety of different information in the trajectory data to construct a feature vector, a comprehensive behavior analysis perspective can be provided. Specifically, the average speed and angle mainly focus on dynamic changes, while the average density and loop coefficient emphasize the properties of space and path efficiency, so that this method can more comprehensively capture the behavior characteristics of pedestrians or non-motor vehicles and improve the accuracy of division.

[0056] Furthermore, the radius R is obtained by the sampling interval and average speed of the trajectory points, and R is expressed by the following formula:

[0057] R=α*Δt*v

[0058] Among them, α is the adjustment coefficient; Δt is the sampling interval of the trajectory points; v is the average speed.

[0059] As an exemplary illustration, the adjustment coefficient α is set to 2.

[0060] Furthermore, the step of calculating the characteristic distance of the trajectory point includes: calculating the cosine similarity between the characteristic vectors of the past trajectory and the future trajectory to obtain the characteristic distance of the trajectory point.

[0061] Specifically, if the past and future trajectories of a trajectory point have significantly different features, their cosine similarity will be lower, that is, the feature distance is larger. This method uses the feature distance as the basis for judging whether a trajectory point has a change in movement state. It not only provides an intuitive way to identify and quantify behavioral changes, but also because the calculation of cosine similarity only depends on the direction of the vector rather than its magnitude, the method has good robustness when dealing with changes at different speeds or in the same behavioral mode.

[0062] In an optional embodiment, the step of obtaining the trajectory point of the movement mode change based on the characteristic distance of the trajectory point includes: fitting the characteristic distance of each trajectory point into a curve FitDist(p i ), where p i represents the i-th trajectory point; based on the curve FitDist(p i ) selects the trajectory points that meet the mobile mode change condition to obtain the trajectory points of the mobile mode change; wherein the expression of the mobile mode change condition is as follows:

[0063] FitDist′(p i-1 )>0andFitDist′(p i )=0andFitDist′(p i )<0

[0064] Among them, FitDist′(p i ) is the function FitDist(p i ) in p i The derivative of .

[0065] In this embodiment, the characteristic distance curve of the trajectory is fitted and its extreme value is used to adaptively find the change point. Specifically, the characteristic distance of the trajectory point is fitted into a curve and its derivative is analyzed to identify the change point of the movement mode. This method can accurately identify the key turning points in the characteristic distance of the trajectory data and determine whether the movement mode has changed, which is beneficial to enhance adaptability and does not rely on a preset threshold, thereby enhancing robustness.

[0066] In an optional embodiment, the step of dividing the trajectory data into different sub-trajectories based on the trajectory points where the movement mode changes, and determining whether the sub-trajectories are in a moving or stationary state comprises: calculating the average density f3(s i ) and loop coefficient f4(s i ), and based on the preset value, it is judged as moving or stationary state, where the expression of the judgment method is as follows:

[0067] f3(s i )>N s andf4(s i ) <C s

[0068] Among them, N s is the preset density threshold, c s is the preset loop coefficient threshold. If the expression is satisfied, the sub-trajectory is judged to be in a stationary state; otherwise, the sub-trajectory is judged to be in a moving state.

[0069] Specifically, if pedestrians or non-motor vehicles are in a stopped state, they must meet the requirements of high density and directional stability. If a pedestrian or non-motor vehicle is in a stopped or extremely slow moving state, the trajectory points generated by it will be closely clustered on the map; directional stability refers to the continuity of the trajectory direction of the individual during the movement; directional stability can eliminate the misjudgment of the moving state caused by high density due to U-turns or similar moving trajectories.

[0070] As an example, the density threshold N s Set to 4.

[0071] As another example, if the moving direction deviates by no more than 45 degrees, the original moving direction is maintained, and the loop coefficient threshold C s Set to

[0072] In this embodiment, a high density threshold and a directional stability threshold are used to determine whether an individual is in a moving or stationary state. This method can improve the accuracy of location data analysis while ensuring that the calculation is simple and fast, especially in high-density environments such as traffic congestion or crowded areas; and by considering directional continuity, the possibility of misjudgment is reduced, providing a more comprehensive understanding of behavior.

[0073] Example 2

[0074] This embodiment proposes a pedestrian and non-motor vehicle trajectory segmentation system based on change point detection, and applies the pedestrian and non-motor vehicle trajectory segmentation method based on change point detection proposed in Example 1. Figure 2 , which is an architecture diagram of a pedestrian and non-motor vehicle trajectory segmentation system based on change point detection in this embodiment.

[0075] This embodiment proposes a pedestrian and non-motor vehicle trajectory segmentation system based on change point detection, including:

[0076] A trajectory data preprocessing module is used to preprocess the acquired trajectory data;

[0077] A feature distance calculation module is used to calculate the feature distance of each trajectory point in the trajectory data;

[0078] The trajectory segmentation module is used to calculate the trajectory points where the movement mode changes based on the characteristic distance of each trajectory point, and divide the trajectory data into different sub-trajectories on this basis, and judge them as moving or stationary states, merge sub-trajectories in the same state, and obtain the pedestrian and non-motor vehicle trajectory segmentation results.

[0079] It can be understood that the system of this embodiment corresponds to the method of the above-mentioned embodiment 1, and the options in the above-mentioned embodiment 1 are also applicable to this embodiment, so they will not be described repeatedly here.

[0080] Example 3

[0081] This embodiment proposes a computer device, including a memory and a processor, wherein the memory stores computer-readable instructions, wherein when the computer-readable instructions are executed by the processor, the processor executes the steps of the pedestrian and non-motor vehicle trajectory segmentation method based on change point detection proposed in Example 1.

[0082] The terms in the drawings are for illustrative purposes only and should not be construed as limiting this patent;

[0083] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the embodiments here. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the claims of the present invention.

Claims

1. A pedestrian and non-motor vehicle trajectory segmentation method based on change point detection, characterized in that: The following steps are involved: Acquire trajectory data including a number of trajectory points and perform preprocessing; Calculating a characteristic distance of each trajectory point in the trajectory data; Obtaining a trajectory point where the movement mode changes based on the characteristic distance of the trajectory point; The trajectory data is divided into different sub-trajectories based on the trajectory points where the movement mode changes, and the sub-trajectories are judged to be in a moving or stationary state, and the sub-trajectories in the same state are merged to obtain a pedestrian and non-motor vehicle trajectory segmentation result.

2. The pedestrian and non-motor vehicle trajectory segmentation method based on change point detection according to claim 1 is characterized in that: The track points include latitude, longitude and time; The preprocessing step includes: analyzing the trajectory data and individual trajectory points and removing trajectory points that do not meet preset requirements.

3. The pedestrian and non-motor vehicle trajectory segmentation method based on change point detection according to claim 1 is characterized in that: The step of calculating the characteristic distance of each trajectory point in the trajectory data comprises: For any trajectory point, the trajectory data of a certain distance before and after the trajectory point is selected as its past trajectory and future trajectory respectively; The feature vectors of the past trajectory and the future trajectory are calculated, and the feature distance of any trajectory point is calculated based on the feature vectors of the past trajectory and the future trajectory.

4. The pedestrian and non-motor vehicle trajectory segmentation method based on change point detection according to claim 3 is characterized in that: The feature vector includes the average speed, average angle, average density and loop coefficient of all trajectory points in the trajectory data; the average density includes the average number of neighbor points of the trajectory point; the loop coefficient is the ratio of the straight-line distance of the trajectory data to the actual movement distance; the expressions of the average density and the loop coefficient are as follows: Among them, f3(s i ) is the average density, where N(p i ) represents the i-th trajectory point p i The number of neighboring points within the radius R; l represents the number of trajectory points contained in the trajectory data; f4(s i ) is the loop coefficient, eu(·) represents the Euclidean distance calculation function, and n represents the number of trajectory points.

5. The pedestrian and non-motor vehicle trajectory segmentation method based on change point detection according to claim 4 is characterized in that: The radius R range is obtained by the sampling interval and average speed of the trajectory points, and R is expressed by the following formula: R=α*Δt*v Among them, α is the adjustment coefficient; Δt is the sampling interval of the trajectory points; v is the average speed.

6. The pedestrian and non-motor vehicle trajectory segmentation method based on change point detection according to claim 3 is characterized in that: The step of calculating the characteristic distance of any trajectory point includes: calculating the cosine similarity between the characteristic vectors of the past trajectory and the future trajectory to obtain the characteristic distance of the trajectory point.

7. The pedestrian and non-motor vehicle trajectory segmentation method based on change point detection according to claim 1 is characterized in that: The step of obtaining the trajectory point of the movement mode change based on the characteristic distance of the trajectory point comprises: fitting the characteristic distance of each trajectory point into a curve FitDist(p i ), where p i represents the i-th trajectory point; based on the curve FitDist(p i ) selects the trajectory points that meet the mobile mode change condition to obtain the trajectory points of the mobile mode change; wherein the expression of the mobile mode change condition is as follows: FitDist′(p i-1 )>0andFitDist′(p i )=0andFitDist′(p i )<0 Among them, FitDist′(p i ) is the function FitDist(p i ) in p i The derivative of .

8. The pedestrian and non-motor vehicle trajectory segmentation method based on change point detection according to claim 1, characterized in that: The step of dividing the trajectory data into different sub-trajectories based on the trajectory points where the movement mode changes, and determining whether the sub-trajectories are in a moving or stationary state comprises: calculating the average density f3(s i ) and loop coefficient f4(s i ), and determines whether it is a moving or stationary state based on a preset determination condition, wherein the expression of the preset determination condition is as follows: f3(s i )>N s and f4(s i )<C s Among them, N s is the preset density threshold, C s is a preset loop coefficient threshold. If the preset judgment condition is met, the sub-track is judged to be in a stationary state; otherwise, the sub-track is judged to be in a moving state.

9. A pedestrian and non-motor vehicle trajectory segmentation system based on change point detection, applied to the pedestrian and non-motor vehicle trajectory segmentation method according to any one of claims 1 to 8, characterized in that: The system comprises: A trajectory data preprocessing module is used to preprocess the acquired trajectory data; A feature distance calculation module is used to calculate the feature distance of each trajectory point in the trajectory data; The trajectory segmentation module is used to calculate the trajectory points where the movement mode changes based on the characteristic distance of each trajectory point, and divide the trajectory data into different sub-trajectories on this basis, and judge them as moving or stationary states, merge sub-trajectories in the same state, and obtain the pedestrian and non-motor vehicle trajectory segmentation results.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the pedestrian and non-motor vehicle trajectory segmentation method based on change point detection according to any one of claims 1 to 8 is implemented.

Citation Information

Patent Citations

  • Urban intersection automatic identification method based on low precision space-time trajectory data

    CN105788273A

  • Traffic trajectory clustering similarity calculation method based on shape factor adjustment

    CN110443311A

  • Method and system for improving data quality of riding track of non-motor vehicle

    CN115310522A

  • Spatio-temporal DP method based on ship trajectory characteristic point extraction

    US20230118438A1