Smart-Phone Trace Segmentation for Trip Information Extraction

Trajectory information recognition is an efficient and convenient approach for trip information obtaining. The trajectory segmentation is a pivotal step to group data into travel parts and activity parts. This paper presents a study on a novel fast search clustering method generalized into the trajectory segmentation. The clustering algorithm was extended and the selection of the cut-off distance was discussed. A comparison among four segmentation methods including two empirical and two clustering methods were conducted to examine the effect of them. The result shows that the two clustering methods achieve lower segment error rate, point error rate, and miss rate than the empirical methods. The fast search algorithm works with the three error rates below 0.03, which can provide reliable samples for information recognition.