Abstract: Extracting and mining information from vehicle trajectories based on clustering algorithm is of great significance to many applications such as high-precision lane information extraction and update, road congestion analysis and treatment, user trip route planning and recommendation. In view of the shortcomings of the existing clustering algorithms, a vehicle trajectory points clustering method based on kernel distance is proposed. Firstly, we give the definition of vehicle trajectory point. Then the geometric characteristics of vehicle trajectories and the requirements of trajectory clustering are analyzed. Thus, we derive the calculation process of kernel distance based on the definition of kernel function. To clarify the kernel Density?Based Spatial Clustering of Application with Noise (DBSCAN) algorithm, the concepts of kernel neighborhood and core object in DBSCAN algorithm are redefined. Finally, the trajectories of Zhengzhou taxi are used to verify the efficiency of the algorithm. Experiments show that the clustering results had significant advantages in reducing the number of parameters, symmetrical distribution along the central line of road, reducing computing time, extracting long clusters and other aspects. It can effectively cluster the directed trajectory points.