关键词: 轨迹压缩, 行驶特征, 自信息量, 马尔可夫序列

Abstract: The popularity of mobile terminals and the development of global positioning system (GPS) produce mass of trajectory data. Based on the data, a lot of location-based services (LBS) provide services for people. However, the increment of trajectory data brings many challenges: huge data volume, long query latency, hard to analyze and data  redundancy. Hence the trajectory compression plays an important role for better LBS. This paper proposes a trajectory compression method which fully considers the moving features and handles trajectory as Markov sequence. The moving features include speed, direction and position. Gaussian distributionis used to model speed change, direction change and position distance, so that the status of the next point can be predicated well based on aforementioned information. According to the prediction accuracy, every point in trajectory is given a conditional self-information value. A compressed trajectory is generated by picking those points that satisfy the user-predefined value. Finally, this paper verifies the performance of the proposed compression method through a number of experiments on real datasets.

Key words: trajectory compression, moving features, self-information, Markov sequence