Streaming Trajectory Segmentation Based on Stay-Point Detection

Trajectory segmentation is essential for various location-based applications. Most of the existing trajectory segmentation algorithms are designed for offline data, thus cannot segment trajectory streams efficiently. In this paper, we propose a novel and efficient streaming trajectory segmentation algorithm based on stay point detection. Our algorithm can dynamically update in real time based on input points, eliminating the need for scanning the entire trajectory. To enhance its suitability for streaming scenarios, we introduce a well-designed grid index and identify three areas based on it for calculation pruning. We conduct extensive experiments using three diverse datasets from different domains (ranging from taxis to trucks to pedestrians), which verifies the efficiency of our method.