An overview on trajectory outlier detection

作者：Meng, Fanrong; Yuan, Guan^*; Lv, Shaoqian; Wang, Zhixiao^*; Xia, Shixiong

来源：Artificial Intelligence Review, 2019, 52(4): 2437-2456.

DOI：10.1007/s10462-018-9619-1

• Outlier detection
• Moving object data mining
• Trajectory
• Spatial-temporal data

The task of trajectory outlier detection is to discover trajectories or their segments which differ substantially from or are inconsistent with the remaining set. In this paper, we make an overview on trajectory outlier detection algorithms from three aspects. Firstly, algorithms considering multi-attribute. In this kind of algorithms, as many key attributes as possible, such as speed, direction, position, time, are explored to represent the original trajectory and to compare with the others. Secondly, suitable distance metric. Many researches try to find or develop suitable distance metric which can measure the divergence between trajectories effectively and reliably. Thirdly, other studies attempt to improve existing algorithms to find outliers with less time and space complexity, and even more reliable. In this paper, we survey and summarize some classic trajectory outlier detection algorithms. In order to provide an overview, we analyze their features from the three dimensions above and discuss their benefits and shortcomings. It is hope that this review will serve as the steppingstone for those interested in advancing moving object outlier detection.

• 出版日期2019-12
• 单位中国矿业大学（北京）; 桂林电子科技大学