Emergency speed change behavior identification method based on track data of electric bicycle

Abstract

The invention provides a sudden speed change behavior recognition method based on track data of an electric bicycle. The method comprises the following steps: 1) track data acquisition and cleaning: collecting the track data and cleaning missing and drifting data; 2) map matching: accurately mapping the trajectory data to an urban road network; 3) identifying the rapid speed change behaviors: dividing the rapid speed change behaviors into rapid acceleration, rapid deceleration, rapid starting and rapid braking by utilizing track kinematics characteristics, and extracting an acceleration threshold value through 85% or 15% quantile in statistics to calibrate a rapid speed change point so as to identify the rapid speed change behaviors; and 4) road network risk level division: based on the shared electric bicycle track data, counting the occurrence frequency of the sudden speed change behavior of each road section in one day, dividing risk levels according to the occurrence frequency, and displaying the road network risk level on a GIS map platform. According to the invention, the rapid speed change behavior is identified through the track kinematics features, and the precision and efficiency of rapid speed change detection are effectively improved.

Description

Technical Field

[0001] The invention belongs to the technical field of intelligent traffic information processing, and more specifically relates to a method for identifying sudden speed change behavior based on electric bicycle trajectory data. Background Art

[0002] In recent years, electric bicycles have become the preferred means of transportation for residents within a certain distance due to their advantages such as speed, convenience, and affordability. However, the rapid growth in the use of electric bicycles has also caused safety issues. Studies have found that from 2016 to 2020, the annual growth rate of traffic collisions involving electric bicycles in urban areas in China was 17.2%, ranking first among all urban transportation tools. Most electric bicycle collision accidents are closely related to risky riding behaviors. Accurately identifying such behaviors is an important part of strengthening the traffic safety management of electric bicycles.

[0003] Among the many risky riding behaviors, sudden acceleration and braking behaviors have been used as a risk quantification indicator to replace traffic accidents and conflict risks. The key to identifying sudden acceleration and braking behaviors is to obtain the corresponding acceleration threshold. At present, the method of obtaining the acceleration threshold of sudden acceleration and braking of electric bicycles mainly relies on experimental methods. By artificially creating sudden acceleration and braking events under experimental conditions, the acceleration threshold parameters are obtained using inertial measurement units (IMUs), brake sensors, GPS locators, etc. However, due to the influence of vehicle kinematic parameters, driving style of the experimental group, data sampling frequency and other conditions, taking sudden braking as an example, the acceleration threshold value ranges from -0.85m / s 2 To -3.40m / s 2 The huge difference in thresholds, experimental time, and economic costs make it difficult to apply in practice. In addition, both sudden acceleration and sudden braking belong to the category of sudden speed change. According to the moving and stationary state of the vehicle, sudden speed change can actually be divided into four behaviors: sudden acceleration, sudden deceleration, sudden start, and sudden braking. Figure 2 At present, there is no method that can realize the recognition of four kinds of rapid speed change behaviors through a set of technologies and improve the accuracy and applicability of acceleration thresholds at the same time. Summary of the invention

[0004] The technical problem to be solved by the present invention is that, in view of the above-mentioned deficiencies of the prior art, a method for identifying sudden speed changes based on electric bicycle trajectory data is proposed, which is characterized in that the acceleration threshold of sudden speed changes is not obtained by experimental methods, but the electric bicycle trajectory data is used to quickly and accurately identify sudden speed changes through an integrated technical solution of "data cleaning-map matching-behavior identification-risk quantification". At the same time, with the help of emerging data platforms such as shared electric bicycles and takeaway electric bicycles, city-level and all-weather sudden speed change behavior detection can be achieved, the road risk level can be quantified, and technical support can be provided for traffic safety management.

[0005] In order to achieve the above purpose, the present invention adopts the following technical solutions:

[0006] Step 1: Acquire and clean the trajectory data of electric bicycles, mainly cleaning the trajectory data with missing data and drift;

[0007] Step 2: Obtain road network data, arrange the trajectory subchains and trajectory points in the subchains in ascending time order, implement map matching, and accurately map the trajectory chains to the urban road network data;

[0008] Step 3: Use trajectory data to identify sudden speed change behavior;

[0009] Step 4: Classification of road network risk levels.

[0010] According to a method for identifying sudden speed change behavior based on electric bicycle trajectory data provided by the present invention, the step 1 of acquiring and cleaning the electric bicycle trajectory data includes:

[0011] Step 1.1: The electric bicycle trajectory data comes from the anonymized data of the shared electric bicycle operators, including order number, trajectory timestamp, and longitude and latitude coordinates.

[0012] Step 1.2: Use the box plot method to clean up missing points and drift points in the trajectory data. Missing points and drift points refer to: due to factors such as failure of the electric bicycle GPS sensor equipment or obstruction by high-rise buildings in the city, the trajectory points of the vehicle during operation will have data anomalies, which are manifested as data points that deviate too much from the true position (drift point) or are lost (missing point).

[0013] According to a method for identifying sudden speed change behavior based on electric bicycle trajectory data provided by the present invention, the map matching in step 2 includes:

[0014] Step 2.1: Road network data preparation.

[0015] Step 2.2: Arrange the trajectory subchains and the trajectory points in the subchains in ascending time order.

[0016] Step 2.3: Implement map matching. After map matching, each trajectory point will have five additional attributes, namely, segment ID, road grade, segment length, one-way and two-way identification, and offset.

[0017] Step 2.4: Delete invalid matching results.

[0018] According to a method for identifying a sudden speed change behavior based on electric bicycle trajectory data provided by the present invention, the step 3 of identifying the sudden speed change behavior includes:

[0019] Step 3.1: Arrange the trajectory points in the trajectory subchain in ascending time order and calculate the acceleration between adjacent trajectory points.

[0020] Step 3.2: Use the ST-DBSCAN algorithm to identify the trajectory subchain T acc The stopping point interval L clu , used to divide the moving and stationary states.

[0021] Step 3.3: Calibrate the jth stop point set clu j The starting point and ending point of Store the location index set D cs , D ce When the trajectory point Acceleration Or track point When the subsequent point is not a stop point, index forward and backward to the threshold a respectively thr1 、a thr2 , calibrate the braking start point and start end point Store the location index set D me , D ms When the stop point starts at the starting point of the trajectory subchain Or end at the end point of the track subchain When the brake start point and start end point calibration are not performed.

[0022] Step 3.4: Calibrate the emergency start state interval L ac , when the stop point set clu j 、clu j+1 When discontinuous, the emergency start point is located in the position index interval When the stop point set clu j 、clu j+1 When continuous, the emergency start point is located at the stop point set clu j+1 Position index in Store the result in L ac . Calibrate the emergency braking state interval L de , when the stop point set clu j 、clu j+1 When discontinuous, the emergency brake point is located in the position index interval When the stop point set clu j 、clu j+1 When continuous, the emergency braking point is located at the stop point set clu j+1 Position index in Store the result in L de .

[0023] Step 3.5: Extract the cycling trajectory dataset Movement, emergency start, emergency braking state range Calculate the thresholds for rapid acceleration, rapid start, rapid deceleration, and rapid stop based on the 85% or 15% percentile value of acceleration When the trajectory point P in the three state intervals i Acceleration i Corresponding to greater than or less than , mark it as a sharp speed change point.

[0024] Step 3.6: Starting from the first sharp speed change point, if the trajectory subchain T acc Movement, emergency start, emergency brake state range If there is at least one trajectory point that is a sudden speed change point, it is recorded as one sudden speed change behavior.

[0025] According to a method for identifying sudden speed change behavior based on electric bicycle trajectory data provided by the present invention, the road network risk level division in step 4 includes:

[0026] Step 4.1: Based on the shared electric bicycle trajectory data, count the frequency of sudden speed changes on various sections of Yizhuang roads within a day, divide the severity of sudden speed changes with appropriate frequency panes, and display the road network risk level based on the GIS map platform.

[0027] The method for identifying sudden speed change behavior based on electric bicycle trajectory data provided by the present invention has at least the following beneficial effects compared with the prior art:

[0028] (1) Strong adaptability to traffic scenarios. The present invention does not rely on specific experimental conditions, has strong adaptability to traffic scenarios, and can work in a variety of different traffic scenarios and conditions.

[0029] (2) No need to experiment to obtain the threshold. The present invention does not rely on experimental methods to obtain the acceleration threshold of the sudden speed change behavior, but directly uses trajectory data analysis, which reduces the time and economic cost of the experiment, and at the same time reduces the measurement errors caused by vehicle kinematic parameters, experimental group driving style, data sampling frequency and other conditions.

[0030] (3) All-weather, city-level detection. With the help of data platforms such as shared electric bicycles, the present invention can achieve city-level, all-weather detection of sudden speed changes, providing more extensive data support for traffic safety management.

[0031] (4) Integrated technical solution. Through the integrated technical solution of "data cleaning-map matching-behavior identification-risk quantification", the present invention realizes the full process automation from data acquisition to risk assessment, which helps to improve the efficiency and effectiveness of traffic safety management. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 A flowchart of a method for identifying a sudden speed change behavior based on electric bicycle trajectory data;

[0033] Figure 2 Schematic diagram of four types of rapid speed change behaviors;

[0034] Figure 3 Schematic diagram of trajectory drift points;

[0035] Figure 4 Schematic diagram of missing trajectory points;

[0036] Figure 5 This is a schematic diagram of the braking, parking and starting state points considering the acceleration value;

[0037] Figure 6 This is a schematic diagram of the emergency start state interval;

[0038] Figure 7 This is a schematic diagram of the emergency braking state range;

[0039] Figure 8 This is a schematic diagram of the frequency distribution of sudden acceleration behavior in the road network;

[0040] Fig. 9 This is a schematic diagram of the frequency distribution of sudden deceleration behavior in the road network;

[0041] Fig.10 This is a schematic diagram of the frequency distribution of emergency start behavior in the road network;

[0042] Fig.11 This is a schematic diagram of the frequency distribution of emergency braking behavior in the road network;

[0043] Fig.12 This is a schematic diagram of Beijing Yizhuang Experimental Area;

[0044] Fig.13 Schematic diagram of the experimental vehicle;

[0045] Fig.14 Schematic diagram of the effectiveness of rapid speed change behavior recognition. DETAILED DESCRIPTION

[0046] In order to make the purpose, technical solution and advantages of the present invention clearer and more specific, the present invention is further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0047] Please see attached Figure 1 The present invention provides a method for identifying a sudden speed change behavior based on electric bicycle trajectory data, comprising the following steps:

[0048] Step 1: Acquire and clean the trajectory data of electric bicycles, mainly cleaning the trajectory data with missing data and drift;

[0049] Step 2: Map matching, accurately mapping the trajectory chain to the urban road network;

[0050] Step 3: Using trajectory data, obtain the acceleration threshold of the sudden speed change behavior;

[0051] Step 4: Classification of road network risk levels.

[0052] The following are application examples of the present invention.

[0053] The sudden speed change behavior is identified based on the trajectory data of shared electric bicycles in Yizhuang area of ​​Daxing District, Beijing.

[0054] (1) Electric bicycle trajectory data acquisition and cleaning

[0055] 1) Data acquisition: The electric bicycle trajectory data comes from the desensitized data of shared electric bicycle operators, including order number, trajectory timestamp, and longitude and latitude coordinates.

[0056] 2) Parameter calculation: Sort the n trajectory points in the trajectory chain in ascending order of time, define the trajectory switch lock point, and set the current trajectory point and the previous trajectory point to be P respectively. i , P i-1 , calculate P i , P i-1 The distance between i-1,i , speed v i-1,i , time t i-1,i , where i is the time sorting identifier of the trajectory point.

[0057] 3) Data cleaning: Using the box plot method, design a data missing point and drift point cleaning strategy that considers the “time threshold + speed threshold”. In the box plot, when the value is lower than Q L -1.5IQR or above U +1.5IQR will be identified as an outlier, where Q L and Q U Represent the lower quartile and upper quartile respectively, and IQR is the interquartile range (IQR = Q U -Q L ). Figure 3-4 As shown, for the drift point, when d i-1,i >v thr t i-1,i When , delete the trajectory point P i Among them, v thr is the upper limit of the permissible speed. For the missing point, when t i-1,i >t thr When , the original trajectory chain is truncated to generate a new trajectory subchain. thrFor a trajectory subchain, when the number of trajectory points exceeds 10 consecutively, it will be retained as a trajectory subchain.

[0058] (2) Map Matching

[0059] 1) Road network preparation: Road network data is obtained through the open source map platform OSM (OpenStreet Map), including road ID, road grade, section length, one-way and two-way signs, etc. In order to ensure that the extracted trajectory chains are all located in the urban road network, the internal roads of residential areas, parks, and units are eliminated, and four types of roads are retained: expressway auxiliary roads, main roads, secondary roads, and branch roads.

[0060] 2) Data index: Arrange the trajectory subchains T in ascending order according to the time series mm and trajectory points in the subchain.

[0061] 3) Map matching: The present invention adopts an interactive voting-based map matching (IVMM) algorithm and improves the algorithm by considering the operation characteristics of electric bicycles. First, the road topology is changed from a directed graph to an undirected graph to avoid situations where the reverse trajectory chain cannot be matched or is incorrectly matched to the opposite lane, ensuring that the reverse behavior trajectory chain is correctly matched. Second, an offset field is added (i.e., the trajectory point P i With section R k Projection distance O of the road centerline ik ), used to extract the trajectory chain within the road red line. Again, when calibrating the parameters, the maximum number of candidate road sections α = 5; the neighborhood query radius r = 150m; when G l <10s, the distance weighted function network parameter β=50, when G l ≥10s, β=200, where G l is the mode value of the time interval set between two adjacent trajectory points in the lth trajectory chain. Finally, after map matching, each trajectory point will add five attributes, namely, segment ID, road grade, segment length, one-way and two-way identification, and offset.

[0062] 4) Result optimization: Extract trajectory chains within different levels of roads according to the road red line range. For one-way roads, the offsets of expressway auxiliary roads, main roads, secondary roads, and branches are 30m, 25m, 20m, and 15m respectively. For two-way roads, the offsets of main roads, secondary roads, and branches are 35m, 30m, and 20m respectively. Delete trajectory chains that are not within the offset buffer range.

[0063] (3) Rapid speed change behavior recognition

[0064] 1) Data index: The trajectory subchain T acc Medium acc The trajectory points are sorted in ascending order of time, and the current trajectory point and the previous trajectory point are defined as Pi , P i-1 .

[0065] 2) Acceleration calculation: using P i-1 , P i The point velocity v i-1 、v i Difference Δv i-1,i , calculate the trajectory point P i The acceleration a i , a i =Δv i-1,i / t i-1,i .

[0066] 3) Stop point identification: The trajectory subchain T acc , spatial neighborhood Eps1, temporal neighborhood Eps2, and minimum number of neighborhood points MinPts are introduced into the ST-DBSCAN algorithm to identify the stop point interval L during riding. clu , used to divide the moving and stationary states.

[0067] 4) Braking, parking and starting status point identification: as shown in the attached Figure 5 As shown, first, calibrate the set of stopping points clu j The starting point position End point position Store the stop start point position index set D respectively cs , stop end point position index set D ce When the trajectory point Acceleration When, from Start indexing forward to a thr1 When <a<0, stop and mark the position of the trajectory point as the braking starting point Store the braking start point position index set D me . Secondly, when the trajectory point When the subsequent point is not a stop point, Start indexing backwards to 0<a<a thr2 Stop when the track point is marked as the start end point And store it in the start and end point position index set D ms In this process, when the stop point starts at the trajectory subchain T acc Starting position When the stop point ends at the trajectory subchain T, the braking start point recognition is not performed; acc End point When the startup end point recognition is not performed. thr1 、a thr2 Take -0.02m / s respectively 2 , 0.02m / s2 .

[0068] 5) Emergency start and emergency brake status interval identification: as shown in the attached Figure 6 As shown, during the startup process, when the parking point set clu j 、clu j+1 When discontinuous, the emergency start point is included in each start state position index interval When the stop point set clu j 、clu j+1 When continuous, the emergency start point is included in the stop point set clu j+1 The starting point position index Store all possible emergency start positions in the emergency start state interval L ac . As attached Figure 7 As shown, during the braking process, when the stop point set clu j 、clu j+1 When discontinuous, the emergency braking point is included in each braking state position index interval When the stop point set clu j 、clu j+1 When continuous, the emergency braking point is included in the stop point set clu j+1 Position index in All possible positions of emergency brakes are stored in the emergency brake state interval L de .

[0069] 6) Rapid speed change point identification: using the rapid start state interval L ac , Emergency braking state interval L de , Stop point interval L clu , extract the riding trajectory dataset Movement, emergency start, emergency braking state range Calculate the thresholds for rapid acceleration, rapid start, rapid deceleration, and rapid stop based on the 85% or 15% percentile value of acceleration Compare sports, emergency start, emergency brake Trajectory point P in the three state intervals i Acceleration i With the rapid change threshold Size. Then the trajectory point P i Identify as a rapid acceleration point; if Then the trajectory point P i Identify as a sudden deceleration point; if Then the trajectory point P i Identify as an emergency start point; if Then the trajectory point P i It is identified as an emergency braking point; otherwise it is a non-sudden speed change point.

[0070] 7) Rapid speed change behavior identification: Starting from the first rapid speed increase point, if the trajectory subchain T acc Movement status range If there is at least one trajectory point in the trajectory that is a sudden acceleration point, it is recorded as one sudden acceleration behavior. Starting from the first sudden reduction point, if the trajectory subchain T acc Movement status range If there is at least one trajectory point in the trajectory that is a sudden deceleration point, it is recorded as one sudden deceleration behavior. Starting from the first sudden start point, if the trajectory subchain T acc Emergency start state range If there is at least one trajectory point in the trajectory that is an emergency start point, it is recorded as an emergency start behavior. Starting from the first emergency braking point, if the trajectory subchain T acc Emergency brake state range If there is at least one trajectory point that is an emergency braking point, it is recorded as one emergency braking behavior.

[0071] (4) Classification of road network risk levels

[0072] 1) Based on the shared electric bicycle trajectory data, the frequency of sudden speed changes on various sections of Yizhuang roads in one day was counted, and three levels were divided with appropriate frequency panes. The sections were marked with green, red and black colors on the GIS map platform to indicate the severity of the sudden speed change behavior, so as to observe the risk level at the road network level, as shown in the attached figure. Figure 8-Figure 11 shown.

[0073] In order to verify the method for identifying sudden speed change behavior based on electric bicycle trajectory data proposed in the present invention, a real vehicle test of sudden speed change behavior was carried out in Yizhuang area. The specific steps for verifying the effectiveness of sudden speed change behavior recognition are as follows:

[0074] 1) Actual vehicle test: Please refer to the attached Fig.12 The experimental vehicle uses an electric bicycle with similar kinematic parameters to the shared electric bicycle as the experimental vehicle, and the sampling time of the GPS locator is set to 1s. Fig.13 As shown. During the experiment, the experimenter actively generated a sudden speed change behavior under the premise of ensuring safety. Video recording was used throughout the experiment to record the time and location of the stop and sudden speed change, which was used as the truth value to verify the effectiveness of the sudden speed change behavior recognition. The entire real vehicle test obtained 32 stops, 16 sudden accelerations, 16 sudden decelerations, 16 sudden starts, and 16 sudden stops.

[0075] 2) Validity verification: The present invention uses the trajectory data with a sampling time of 1s as the basis for data cleaning and map matching, and then samples the trajectory equally to 10s, obtaining a total of 10 sets of trajectory data to verify the validity of the algorithm. The results are shown in the attached figure. Fig.14As shown. First, as the sampling time increases, the sudden speed change threshold gradually decreases, which reflects that the different sudden speed change thresholds obtained in previous studies may be greatly affected by the sampling time. Secondly, when the trajectory sampling interval is within 6s, the recognition effectiveness of most sudden speed change behaviors can reach more than 84.21%, reflecting the effectiveness of the electric bicycle sudden speed change behavior recognition algorithm proposed in the present invention. Finally, when the trajectory sampling interval is greater than 6s, the effectiveness of sudden speed change behavior recognition begins to decay rapidly, reflecting that the trajectory data with a trajectory sampling interval greater than 6s is no longer suitable for sudden speed change behavior recognition.

[0076] Matters not covered by the present invention are known technologies.

[0077] The above embodiments are only for illustrating the technical concept and features of the present invention, and their purpose is to enable people familiar with the technology to understand the content of the present invention and implement it accordingly, and they cannot be used to limit the protection scope of the present invention. Any equivalent changes or modifications made according to the spirit of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for identifying sudden speed change behavior based on electric bicycle trajectory data, characterized in that: According to the kinematic characteristics of the trajectory, the rapid speed change is divided into four behaviors: rapid acceleration, rapid deceleration, rapid start and rapid braking. The identification method includes the following steps: Step S1: acquiring and cleaning electric bicycle trajectory data; Step S2: Obtain road network data, arrange the trajectory subchains and the trajectory points in the subchains in ascending time order, perform map matching, and accurately map the trajectory chains to the urban road network data; Step S3: using the trajectory data to identify the sudden speed change behavior; Step S4: Classification of road network risk levels.

2. The method for identifying sudden speed change behavior based on electric bicycle trajectory data according to claim 1, characterized in that: The process of acquiring and cleaning the trajectory data in step S1 is as follows: Step S11: The electric bicycle trajectory data comes from the desensitized data of the shared electric bicycle operator, including the order number, trajectory timestamp, and longitude and latitude coordinates; Step S12: cleaning missing points and drifting points in the trajectory data; missing points and drifting points refer to the trajectory points of the vehicle during operation that will have data anomalies due to failure of the electric bicycle GPS sensor equipment or obstruction by high-rise buildings in the city, which are manifested as data points that deviate too much from the actual position or are missing.

3. The method for identifying sudden speed change behavior based on electric bicycle trajectory data according to claim 1, characterized in that: The map matching process in step S2 is as follows: Step S21: road network data preparation; Step S22: After map matching, each trajectory point will have five additional attributes, namely, section ID, road grade, section length, one-way and two-way identification, and offset; Step S23: Delete invalid matching results.

4. The method for identifying sudden speed change behavior based on electric bicycle trajectory data according to claim 1, characterized in that: The process of identifying the sudden speed change behavior in step S3 is as follows: Step S31: Arrange the trajectory points P in the trajectory subchain in ascending order of time i , calculate the acceleration between adjacent trajectory points, where i is the time sequence identifier of the trajectory point; Step S32: Using the ST-DBSCAN algorithm to identify the trajectory subchain T acc The stopping point interval L clu , used to divide the motion and stationary states; Step S33: calibrate the jth stop point set clu j The starting and ending points of And store the stop start point and end point position index set D cs , D ce ; is the current trajectory point Acceleration Or track point When the subsequent point is not a stop point, index forward and backward to the threshold a respectively thr1 、a thr2 , to calibrate the braking start point and the start end point And store the braking start point and start end point position index set D me , D ms ; When the stop starts or ends at the start or end point of the track subchain The braking start point or the starting end point is not calibrated; Step S34: calibrate the emergency start and emergency brake state interval L ac , L de ; For L ac , when the stop point set clu j 、clu j+1 When discontinuous, the emergency start point is located in the position index interval When the stop point set clu j 、clu j+1 When continuous, the emergency start point is located at the stop point set clu j+1 Position index in For L de , when the stop point set clu j 、clu j+1 When discontinuous, the emergency brake point is located in the position index interval When the stop point set clu j 、clu j+1 When continuous, the emergency braking point is located at the stop point set clu j+1 Position index in Step S35: Extracting the motion, sudden acceleration, and sudden braking state intervals from the riding trajectory data set Use acceleration quantiles to determine the thresholds for rapid acceleration, rapid start, rapid deceleration, and rapid stop If a trajectory point P in the three state intervals i The acceleration a i Corresponding to more than or less than Then this point is marked as a sharp speed change point; Step S36: Starting from the first rapid speed change point, if the track subchain T acc Movement, emergency start, emergency brake state range If there is at least one trajectory point that is a sudden speed change point, it is recorded as one sudden speed change behavior.

5. The method for identifying sudden speed change behavior based on electric bicycle trajectory data according to claim 1, characterized in that: The process of dividing the road network risk level in step S4 is as follows: Based on the trajectory data of shared electric bicycles, the frequency of sudden speed changes on each road section within a day is counted, and the severity of sudden speed changes is divided by appropriate frequency panes, and the road network risk level is displayed based on the GIS map platform.

Images