Recognition Of Bad Driving Behavior And Analysis Of Spatiotemporal Distribution Based On Vehicle Trajectory

As operating vehicles must be equipped with satellite positioning devices and connected to the monitoring platform,a large amount of GPS-based track data of operating vehicles has been generated.Trajectories not only record the position and state information of vehicles,but also contain a large amount of semantic information,correlation between track segments and a large number of unmined patterns.Therefore,it is of great significance for the management of transportation enterprises and the supervision of related departments to excavate the track data of operating vehicles and analyze the movement pattern and behavior characteristics of operating vehicles.Based on the historical track of 502 freight vehicles,this paper systematically studie’s the track data mining of operating vehicles.According to the track data of operating vehicles,seven kinds of bad driving behavior identification models based on characteristic threshold and a track clustering algorithm based on adaptive radius were proposed.Based on the identified bad driving behavior data,a clustering algorithm,RN-DBSCAN,was proposed to analyze hot spots in road network.The specific research contents of this paper are as follows:(1)Based on the track data of operating vehicles,the data were cleaned and the characteristics such as acceleration,steering acceleration and vehicle speed stability were extracted.Based on the original features and extracted features in the trajectory,seven kinds of bad driving behavior identification models were designed,and the distribution of the identified bad driving behaviors in time and space was statistically analyzed.(2)In order to mine the hot spots of bad driving behavior distribution,this paper proposes a clustering algorithm RN-DBSCAN for hot spots analysis in road network.In order to accurately measure the distance between two events in the road network,the plane road network distance is designed as the similarity calculation method of the algorithm.Based on the bad driving behavior data,the parameter selection of the algorithm is discussed,and the efficiency and the quality of clustering results are compared with DBSCAN algorithm.(3)In order to perform clustering analysis on the track of operating vehicles,the classical track clustering algorithm Traclus was improved in this paper.In the phase of track segmentation,considering the regularity of track distribution in the road network,a simple and reliable algorithm of vehicle track segmentation is proposed.In the stage of clustering,a clustering algorithm based on adaptive radius is proposed.The validity of the algorithm was verified based on 1523 operating vehicle tracks.