Research and Application of Trajectory Stop Point Detection Algorithm for Time Series Clustering

Abstract

In order to solve the problem of low accuracy of sampling irregular tracks, a time series clustering algorithm for detecting stops is proposed. Firstly, based on the data field theory, a hybrid feature density detection method considering temporal and spatial characteristics is designed. Secondly, according to the characteristic that the center density of the stop point is greater than the inlet density, the filtering and refining strategy is used to extract the stop point. In the filtration stage, the time duration and the minimum density threshold are selected as the candidate residence points. The maximum threshold is used to identify the actual residence point in the refining stage. The experimental results show that the proposed method can effectively detect the residence points on the irregular trajectories with higher accuracy and less time consumption than the existing methods.