DOI: 10.3724/SP.J.1004.2013.02131

Acta Automatica Sinica (自动化学报) 2013/39:12 PP.2131-2142

Determination of Real-time Vehicle Driving Status Using HMM

In this paper, we propose a method for determination of vehicle driving status from its time-ordered trajectory data using the hidden Markov model (HMM). Firstly, we take some preprocessing including linear smooth filtering and least square fitting to abandon the trajectory sequences whose lengths are not enough, so as to guarantee the usability of acquired trajectory sequences. Secondly, we extract trajectory direction angle features from the trajectory sequences, and on this basis we propose a direction angle region partition algorithm to generate the observation sequences, which will determine the di®erent trajectory patterns acquired by vehicle real-time various driving status. Finally, we get the optimal HMM model parameters of each trajectory pattern in specific traffic scene by multiple observations based Baum-Welch algorithm, then through matching with the above trained HMM models, we can determine the real-time vehicle driving status. Experiment results demonstrate the effectiveness and stability of this method.

Key words:Video vehicle trajectories, hidden Markov models (HMM), direction angles, driving status, realtime deter-mination

ReleaseDate:2014-07-21 17:04:35

Funds:National Natural Science Foundation of China (41271422), Specialized Research Fund for the Doctoral Program of Higher Education (20132136110002), Natural Science Foundation of Liaoning Province (20102123), Open Foundation of Novel Software Technology of State Key Laboratory (Nanjing University) (KFKT2011B09, KFKT2011B11), Open Foundation of Image Processing and Image Communication Laboratory (Nanjing University of Posts and Telecommunications) of Jiangsu Province (LBEK2010003), and Intelligent Computing and Information Processing, Open Topics of Education Ministry (Xiangtan University) (2011ICIP06)