Vehicle abnormal parking behavior identification method

Abstract

The invention discloses a vehicle abnormal parking behavior identification method, and relates to the technical field of intelligent traffic, and the method comprises the steps: obtaining an original track sequence of a vehicle in a driving process through a vehicle-mounted GPS; filtering the original track sequence according to the driving state fluctuation condition of the vehicle at each track point in the original track sequence to obtain a processed track sequence; extracting a candidate parking section sequence from the processed track sequence, extracting a plurality of candidate parking points according to the position of each track point in the candidate parking section sequence, and clustering the extracted candidate parking points to obtain a clustering center; and when the clustering center is not in the allowed parking area and the distance between the clustering center and the preset driving path exceeds an offset threshold value, determining that the vehicle has an abnormal parking behavior. The accuracy of the trajectory data is effectively improved through data filtering and segmentation processing, and the trajectory data is combined with a preset driving path and an allowed parking area, so that the abnormal parking behavior is efficiently and accurately identified.

Description

Technical Field

[0001] The present application relates to the field of intelligent transportation technology, and in particular to a method for identifying abnormal parking behavior of a vehicle. Background Art

[0002] Abnormal parking behavior of vehicles during driving seriously affects the normal order of road traffic. Random parking outside the permitted parking area will cause traffic jams and even traffic accidents. Therefore, accurately identifying abnormal parking behavior on the road is of great significance for the safe and stable operation of the transportation system.

[0003] Currently, there are methods for identifying abnormal parking behavior through video surveillance, but this method is often limited by the coverage of surveillance cameras and is easily affected by weather, light, and obstructions. The accuracy and reliability of identifying abnormal parking behavior of vehicles are not high. Summary of the invention

[0004] In view of the above problems and technical requirements, this application proposes a method for identifying abnormal parking behavior of vehicles. The technical solution of this application is as follows:

[0005] A method for identifying abnormal parking behavior of a vehicle comprises the following steps:

[0006] The original trajectory sequence of the vehicle during driving is obtained through the vehicle-mounted GPS, and the original trajectory sequence includes multiple trajectory points;

[0007] According to the driving state fluctuation of the vehicle at each trajectory point in the original trajectory sequence, the original trajectory sequence is filtered to obtain a processed trajectory sequence;

[0008] A candidate parking segment sequence is extracted from the processed trajectory sequence, wherein the candidate parking segment sequence includes a plurality of continuous trajectory points, and a driving speed of the vehicle at each trajectory point in the candidate parking segment sequence is less than a low speed threshold;

[0009] Extracting several candidate parking points according to the positions of each trajectory point in the candidate parking segment sequence, and clustering the extracted candidate parking points to obtain cluster centers;

[0010] When the cluster center is not within the permitted parking area and the distance between the cluster center and the predetermined driving path exceeds the deviation threshold, it is determined that the vehicle has abnormal parking behavior.

[0011] A further technical solution is to filter the original trajectory sequence to obtain a processed trajectory sequence including:

[0012] Calculate the driving state information of the vehicle at each trajectory point according to the position and time stamp of each trajectory point in the original trajectory sequence, and the driving state information includes the driving acceleration, driving speed and driving direction of the vehicle;

[0013] For any trajectory point in the original trajectory sequence, when the position of the trajectory point is abnormal, and / or the timestamp of the trajectory point is abnormal, and / or the driving state information of the vehicle at the trajectory point is abnormal, the trajectory point is determined to be an abnormal trajectory point; all abnormal trajectory points in the original trajectory sequence are eliminated to obtain a processed trajectory sequence.

[0014] A further technical solution is to extract a candidate parking segment sequence from the processed trajectory sequence, including:

[0015] All trajectory points where the vehicle's driving speed is less than a low-speed threshold are extracted from the processed trajectory sequence as low-speed trajectory points, and the low-speed trajectory points where the change in the vehicle's driving direction at multiple consecutive low-speed trajectory points exceeds the direction change threshold are filtered out; a candidate parking segment sequence consisting of the remaining multiple consecutive low-speed trajectory points is extracted, and the timestamp interval between adjacent low-speed trajectory points in the candidate parking segment sequence does not exceed the time interval threshold.

[0016] Its further technical solution is:

[0017] Extracting several candidate parking points according to the positions of each trajectory point in the candidate parking segment sequence includes:

[0018] A number of trajectory points whose distances between positions in the candidate parking segment sequence do not exceed a distance threshold are merged into one candidate parking point, and the average value of the positions of the trajectory points before merging is used as the position of the merged candidate parking point.

[0019] A further technical solution is that the method for identifying abnormal parking behavior of a vehicle further includes:

[0020] When it is detected that the extracted candidate parking spot is located in the abnormal parking hotspot area, it is directly determined that the vehicle has abnormal parking behavior.

[0021] A further technical solution is that the method for identifying abnormal parking behavior of a vehicle further includes:

[0022] When abnormal parking behavior of a vehicle is detected, the cluster center is used as the abnormal parking position of the vehicle and added to the historical abnormal parking database. All abnormal parking positions in the historical abnormal parking database are counted and abnormal parking hot spots are determined.

[0023] Its further technical solution is to count all abnormal parking locations in the historical abnormal parking database and determine the abnormal parking hot spots including:

[0024] When the number of times that an area of ​​a predetermined area contains abnormal parking positions in a historical abnormal parking database reaches a number threshold, it is determined that the area belongs to an abnormal parking hotspot area.

[0025] The beneficial technical effects of this application are:

[0026] The present application proposes a method for identifying abnormal parking behavior of vehicles, which is based on the trajectory data of conventional vehicle-mounted GPS, can effectively reduce the dependence on basic equipment such as surveillance video, and is simple and convenient to deploy and maintain. Through data filtering and segmentation processing, the accuracy of the original trajectory sequence can be effectively improved, the influence of noise data generated by positioning errors such as unstable GPS signals can be reduced, the reliability and accuracy of trajectory data can be ensured, and subsequent trajectory data analysis can be based on effective trajectory data, thereby improving the accuracy of abnormal parking behavior identification.

[0027] Candidate parking points are extracted by analyzing the spatial distance and time distance between trajectory points in the candidate parking segment sequence, and the trajectory points in the candidate parking segment sequence are further screened and merged to avoid repeated judgment of the same abnormal parking behavior, which affects the recognition efficiency. Abnormal parking behavior is determined by clustering the merged candidate parking points and comparing the cluster center with the predetermined trajectory path and the allowed parking area. The geographical distribution characteristics of the candidate parking points are fully considered to obtain the parking hotspot area and the reference route and parking rules in the lane are combined to accurately identify abnormal parking behavior, further improving the recognition accuracy. It provides a solid foundation for subsequent driving path planning and parking hotspot analysis, and helps to improve the intelligent level of driving safety management and route planning.

[0028] By batch analyzing abnormal parking behaviors in the historical abnormal parking database, we can use historical data to quickly identify abnormal parking behaviors in candidate parking spots, achieve rapid response to real-time data, and ensure timely identification and feedback of abnormal parking behaviors during actual driving. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 It is a flow chart of a method for identifying abnormal parking behavior of a vehicle.

[0030] Figure 2 This is a flowchart for filtering the original trajectory sequence.

[0031] Figure 3 This is a schematic diagram of clustering candidate parking points. DETAILED DESCRIPTION

[0032] The specific implementation of the present application is further described below in conjunction with the accompanying drawings.

[0033] This application proposes a method for identifying abnormal parking behavior of vehicles, please refer to Figure 1 The flowchart shown in the figure has the following specific steps:

[0034] Step 1: Obtain the original trajectory sequence of the vehicle during driving through the vehicle-mounted GPS, where the original trajectory sequence includes multiple trajectory points.

[0035] The vehicle-mounted GPS can collect the vehicle's track points in real time. Each track point represents the latitude and longitude position of the vehicle at a certain timestamp. The multiple track points obtained constitute the original track sequence. In order to ensure the accuracy of the original data obtained as much as possible, the GPS equipped on the vehicle needs to be selected with the highest possible accuracy. At the same time, the data update frequency is recommended to be within 1 second or less to ensure the real-time nature of the data.

[0036] Step 2: Filter the original trajectory sequence according to the driving state fluctuation of the vehicle at each trajectory point in the original trajectory sequence to obtain a processed trajectory sequence.

[0037] Due to sampling errors caused by unstable GPS signals or delays in the acquisition system, noise data exists in the acquired original trajectory data, including discrete data points with abnormalities such as speed mutation, position drift, direction mutation, and time anomaly. Therefore, it is necessary to filter the original trajectory data first to eliminate the influence of noise data.

[0038] In one embodiment, filtering the original trajectory sequence to obtain a processed trajectory sequence includes: calculating the driving state information of the vehicle at each trajectory point according to the position and timestamp of each trajectory point in the original trajectory sequence, and the driving state information includes the driving acceleration, driving speed and driving direction of the vehicle. The driving direction of the vehicle is generally represented by an azimuth angle, which is the angle of the vehicle's driving direction relative to the north direction clockwise, in degrees, and can be calculated using the latitude and longitude position of the vehicle. The specific calculation process is the prior art and this application will not repeat it.

[0039] During vehicle driving, factors such as sharp turns or U-turns may cause sudden changes in the driving direction of some trajectory points, turning the data into discrete data points. Therefore, to determine whether the discrete data points in the original trajectory sequence are noise data, it is necessary to perform statistical analysis based on the results of a segment of data. A segment of data can be a specified length of data selected based on actual experience or the entire trajectory sequence. By statistically analyzing the standard deviation, variance and other statistical values ​​of a segment of data, the difference of each trajectory point relative to the statistical value is calculated to determine whether it is abnormal, so as to ensure that the characteristics of the overall data are taken into account and avoid mistakenly eliminating some necessary normal driving data. For any trajectory point in the original trajectory sequence, when the position of the trajectory point is abnormal, and / or the timestamp of the trajectory point is abnormal, and / or the driving status information of the vehicle at the trajectory point is abnormal, the trajectory point is determined to be an abnormal trajectory point. All abnormal trajectory points in the original trajectory sequence are eliminated to obtain a processed trajectory sequence.

[0040] In one embodiment, the data can be filtered in order according to the priority of the abnormal judgment conditions to improve the efficiency of data filtering. For example, the driving acceleration of a vehicle is the most important factor in analyzing the driving state of the vehicle. Therefore, the driving acceleration of each track point in the original track sequence is first filtered. When the driving acceleration of the track point is abnormal, the track point is directly removed and no further judgment is performed. For the specific process, please refer to Figure 2 Flowchart shown.

[0041] Step 3: extract a candidate parking segment sequence from the processed trajectory sequence, where the candidate parking segment sequence includes a plurality of continuous trajectory points, and the driving speed of the vehicle at each trajectory point in the candidate parking segment sequence is less than a low speed threshold.

[0042] The processed trajectory sequence is segmented in chronological order, and the processed trajectory sequence is divided into continuous trajectory segments to monitor the vehicle status in a short period of time, thereby providing a data basis for the subsequent identification of abnormal parking behavior. In one embodiment, extracting a candidate parking segment sequence from the processed trajectory sequence includes: extracting all trajectory points of the vehicle whose driving speed is less than the low-speed threshold from the processed trajectory sequence as low-speed trajectory points. Since the inherent defect of GPS positioning error will cause the drift problem of the obtained driving speed and position, it is necessary to extract low-speed trajectory points. In addition, in some cases, the parking behavior of the vehicle is not completely stationary. For example, when the vehicle wants to wait for someone on the roadside, it does not want to stop completely but slows down at a very low driving speed. At this time, the driving speed of the vehicle is not 0, and the trajectory points are not completely overlapped, but this also belongs to abnormal parking behavior, so it is necessary to extract all low-speed trajectory points to cover these special cases to ensure the accuracy of abnormal parking behavior identification. In addition, in order to enter the abnormal parking behavior identification process in advance before the vehicle stops completely, it is necessary to extract the low-speed trajectory points before parking. Among them, the low-speed threshold can be customized according to the actual application.

[0043] The low speed of a vehicle is not necessarily caused by parking. When the vehicle is turning or making a U-turn, the speed will also be low, so these normal driving trajectory points need to be eliminated. Since the change in the driving direction of multiple consecutive trajectory points is large when the vehicle is turning or making a U-turn, the low-speed trajectory points whose change in the driving direction at multiple consecutive low-speed trajectory points exceeds the direction change threshold are filtered out. A candidate parking segment sequence consisting of the remaining multiple consecutive low-speed trajectory points is extracted, and the timestamp interval between adjacent low-speed trajectory points in the candidate parking segment sequence does not exceed the time interval threshold. Among them, the direction change threshold and the time interval threshold can be customized according to the actual application.

[0044] Step 4, extracting several candidate parking points according to the position of each trajectory point in the candidate parking segment sequence, and clustering the extracted candidate parking points to obtain cluster centers. Since each trajectory point in the extracted candidate parking segment sequence is a low-speed trajectory point, and the same abnormal parking behavior may contain multiple low-speed trajectory points, it is necessary to further merge the low-speed trajectory points to obtain candidate parking points to avoid repeated determination of the same abnormal parking point. In one embodiment, extracting several candidate parking points according to the position of each trajectory point in the candidate parking segment sequence includes: merging several trajectory points in the candidate parking segment sequence whose distances between positions do not exceed a distance threshold into one candidate parking point, and using the average value of the positions of each trajectory point before merging as the position of the merged candidate parking point.

[0045] Candidate parking points indicate that there is a high probability that a vehicle will stop at this location. At this time, it is impossible to determine whether the candidate parking point is caused by abnormal parking or normal slow movement of the vehicle, so further clustering analysis is required to determine. Commonly used clustering algorithms such as DBSCAN algorithm, SNN algorithm, OPTICS algorithm, etc. can be used. This application uses the DBSCAN algorithm as an example for illustration. The DBSCAN algorithm is used to cluster candidate parking points by setting the cluster radius and the minimum number of cluster points to obtain several cluster groups. Each cluster group includes multiple candidate parking points and a cluster center, such as Figure 3 As shown, each cluster center represents an abnormal parking hotspot location.

[0046] Step 5: When the cluster center is not within the permitted parking area and the distance between the cluster center and the predetermined driving path exceeds the offset threshold, it is determined that the vehicle has abnormal parking behavior and the cluster center is stored in the historical abnormal parking database.

[0047] The information of permitted parking areas (parking lots, emergency lanes, etc.) and scheduled driving paths (lane center lines, etc.) is the basic road information collected in the monitoring area. The method for calculating the distance between the cluster center and the scheduled driving path based on the basic road information is: decompose the scheduled driving path into multiple line segments, calculate the distance from each cluster center to each line segment, and obtain the shortest distance as the distance between the cluster center and the scheduled driving path. The distance formula from point to line segment can be used to calculate the distance between the cluster center and any line segment of the scheduled driving path. Among them, (x1, y1) and (x2, y2) are the endpoints of the line segment, and (x, y) is the cluster center. The offset threshold can be customized according to the actual road conditions. When the distance between the cluster center and the predetermined driving path exceeds the offset threshold, it means that the vehicle has deviated from the normal driving path. For example, the vehicle has deviated from the center line of the road and moved closer to the roadside. If it is not in the allowed parking area at this time, it means that abnormal parking behavior has occurred.

[0048] In order to improve the efficiency of abnormal parking behavior identification and achieve rapid response to real-time data so as to timely identify and feedback abnormal parking behavior, abnormal parking behavior can be determined by analyzing the abnormal parking hotspot area. In another embodiment, when it is detected that the extracted candidate parking point is located in the abnormal parking hotspot area, it is directly determined that the vehicle has abnormal parking behavior. Abnormal parking behavior can be quickly identified by this method.

[0049] Abnormal parking hotspots refer to areas where abnormal parking behaviors are found multiple times. When abnormal parking behaviors of vehicles are detected, the cluster center is used as the abnormal parking location of the vehicle and added to the historical abnormal parking database. All abnormal parking locations in the historical abnormal parking database are counted and abnormal parking hotspots are determined.

[0050] In one embodiment, a specific method for counting all abnormal parking positions in the historical abnormal parking database and determining abnormal parking hot spots is as follows: when the number of times that an area of ​​a predetermined area contains abnormal parking positions in the historical abnormal parking database reaches a number threshold, the area is determined to be an abnormal parking hot spot. The predetermined area and the number threshold can be customized.

[0051] The above is only a preferred embodiment of the present application, and the present application is not limited to the above embodiments. It is understood that other improvements and changes directly derived or associated by those skilled in the art without departing from the spirit and concept of the present application should be considered to be included in the protection scope of the present application.

Claims

1. A method for identifying abnormal parking behavior of a vehicle, characterized in that: The method for identifying abnormal parking behavior of a vehicle comprises: Acquire an original trajectory sequence of the vehicle during driving through the vehicle-mounted GPS, wherein the original trajectory sequence includes a plurality of trajectory points; The original trajectory sequence is filtered according to the driving state fluctuation of the vehicle at each trajectory point in the original trajectory sequence to obtain a processed trajectory sequence; Extracting a candidate parking segment sequence from the processed trajectory sequence, wherein the candidate parking segment sequence includes a plurality of continuous trajectory points, and the driving speed of the vehicle at each trajectory point in the candidate parking segment sequence is less than a low speed threshold; Extracting a number of candidate parking points according to the positions of each trajectory point in the candidate parking segment sequence, and clustering the extracted candidate parking points to obtain cluster centers; When the cluster center is not within the permitted parking area and the distance between the cluster center and the predetermined driving path exceeds a deviation threshold, it is determined that the vehicle has abnormal parking behavior.

2. The method for identifying abnormal parking behavior of a vehicle according to claim 1, characterized in that: The processed trajectory sequence obtained by filtering the original trajectory sequence includes: Calculating the driving state information of the vehicle at each trajectory point according to the position and time stamp of each trajectory point in the original trajectory sequence, wherein the driving state information includes the driving acceleration, driving speed and driving direction of the vehicle; For any trajectory point in the original trajectory sequence, when the position of the trajectory point is abnormal, and / or the timestamp of the trajectory point is abnormal, and / or the driving state information of the vehicle at the trajectory point is abnormal, the trajectory point is determined to be an abnormal trajectory point; and all abnormal trajectory points in the original trajectory sequence are eliminated to obtain a processed trajectory sequence.

3. The method for identifying abnormal parking behavior of a vehicle according to claim 1, characterized in that: The candidate parking segment sequence is extracted from the processed trajectory sequence, including: All trajectory points where the vehicle's driving speed is less than a low-speed threshold are extracted from the processed trajectory sequence as low-speed trajectory points, and the low-speed trajectory points where the change in the vehicle's driving direction at multiple consecutive low-speed trajectory points exceeds the direction change threshold are filtered out; a candidate parking segment sequence consisting of the remaining multiple consecutive low-speed trajectory points is extracted, and the timestamp interval between adjacent low-speed trajectory points in the candidate parking segment sequence does not exceed the time interval threshold.

4. The method for identifying abnormal parking behavior of a vehicle according to claim 1, characterized in that: Extracting a plurality of candidate parking points according to the positions of each trajectory point in the candidate parking segment sequence includes: A plurality of trajectory points in the candidate parking segment sequence whose distances between positions do not exceed a distance threshold are merged into one candidate parking point, and an average value of the positions of the trajectory points before merging is used as the position of the merged candidate parking point.

5. The method for identifying abnormal parking behavior of a vehicle according to claim 1, characterized in that: The vehicle abnormal parking behavior identification method further includes: When it is detected that the extracted candidate parking spot is located in the abnormal parking hotspot area, it is directly determined that the vehicle has abnormal parking behavior.

6. The method for identifying abnormal parking behavior of a vehicle according to claim 5, characterized in that: The vehicle abnormal parking behavior identification method further includes: When abnormal parking behavior of a vehicle is detected, the cluster center is used as the abnormal parking position of the vehicle and added into a historical abnormal parking database, and all abnormal parking positions in the historical abnormal parking database are counted and abnormal parking hot spots are determined.

7. The method for identifying abnormal parking behavior of a vehicle according to claim 6, characterized in that: Statistics of all abnormal parking locations in the historical abnormal parking database and identification of abnormal parking hot spots include: When the number of times that an area of ​​a predetermined area contains abnormal parking positions in a historical abnormal parking database reaches a number threshold, it is determined that the area belongs to an abnormal parking hotspot area.

Images