On Velocity-Preserving Trajectory Simplification

摘要

Trajectory data plays crucial role in many real-world applications with moving objects. The size of trajectory dataset is always very huge because of high sampling rate. Therefore, it is desired to simplify each trajectory before it is stored and processed. As the result, many trajectory simplification notions have been proposed. However, existing studies on trajectory simplification more or less rely on geometric-preserving manner (e.g., minimizing position-based or direction-based errors). These manners directly avoid effectiveness of velocity in many real-world applications. Actually, the velocity of a moving object is very important in many real-world applications, such as map-matching, mobility prediction, moving pattern mining, etc. In this paper, we propose a novel trajectory simplification, velocity-preserving trajectory simplification (VPTS), which minimize both geometric error and velocity error. We present an efficient algorithm for optimal velocity-preserving trajectory simplification. Through a series of experimental evaluation with real trajectory data, we examine the benefit of our proposed velocity-preserving trajectory simplification.