导航

ACTA AERONAUTICAET ASTRONAUTICA SINICA ›› 2023, Vol. 44 ›› Issue (8): 327281-327281.doi: 10.7527/S1000-6893.2022.27281

• Electronics and Electrical Engineering and Control • Previous Articles     Next Articles

Online classification of target behavior pattern based on spatiotemporal trajectory information

Qiaowen JIANG, Yu LIU(), Ziran DING, Shun SUN, Tao JIAN   

  1. Institute of Information Fusion,Naval Aviation University,Yantai 264001,China
  • Received:2022-04-13 Revised:2022-05-11 Accepted:2022-05-24 Online:2023-04-25 Published:2022-06-08
  • Contact: Yu LIU E-mail:liuyu77360132@126.com
  • Supported by:
    National Natural Science Foundation of China(62022092);Shandong Provincial Colleges and Universities Youth Innovation Technology Support Program(2021KJ005);Foundation Strengthening Program technical area Fund(2021-JCJQ-JJ-0251);Special funds for Taishan Scholars Project(tsqn201909156)

Abstract:

In the early warning and monitoring field, it is very important to predict target behavior patterns in time for situation cognition. Firstly, an Inductive Conformal Multi-class Predictor (ICMP) with outlier detection is proposed to solve the problem that most current classifiers are in low computational efficiency and not sensitive to outliers. Secondly, to overcome the problem that most behavior pattern classifiers only consider the spatial location information of the target and ignore the course and velocity information, a concept of Spatio-Temporal Hausdorff Distance (STHD) is proposed to distinguish behaviors with similar spatial position but different course and velocity effectively. Then, a Directed Spatiotemporal Hausdorff Nearest Neighbor Nonconformity Measure (DSHNN-NCM) function is constructed by combining the idea of KNN and STHD. On this basis, a Sequential Spatiotemporal Hausdorff Nearest Neighbor Inductive Conformal Multi-class Predictor (SSHNN-ICMP) with outlier detector is proposed, which can online learn and predict the frequent behaviors of targets in the early warning and monitoring scenario. Finally, experiments and analyses are carried out in a military simulation scenario and a real civilian scenario. The results show that the proposed algorithm has better accuracy and real-time performance, showing good applicability in early warning and surveillance tasks.

Key words: spatiotemporal trajectory, behavior pattern, Hausdorff distance, conformal predictor, online classification

CLC Number: