CLASSIFICATION OF TRAFFIC EVENTS BASED ON THE SPATIO-TEMPORAL MRF MODEL AND THE BAYESIAN NETWORK
Shunsuke Kamijo, Masao Sakauchi · 2002
In order to support safe and e#cient driving, it is important to classify the behaviors of vehicles and to understand what is going on the tra#c situations. For that purpose, our system employed a vision sensor rather than spot sensors because of its rich information. We then have developed a dedicated vehicle tracking algorithm based on the Spatio-Temporal MRF model which is robust against heavy occlusions even in low-angle images. This algorithm outputs maps that represent distributions of vehicle regions and motions by every image frames. By analyzing time-series variations of these distribution maps, our system extracts observation sequences for tra#c events classification. We then constructed a classification network based on the Bayesian Network which consists of a fusion of deterministic sub-networks and probabilistic sub-networks. Since vehicle behaviors in the tra#c should be restricted by tra#c rules, such behaviors would be classified by deterministic network based on explicitly defined tra#c rules. However, it is di#cult to classify and to understand behaviors such as accidents, near misses, and some other ambiguous behaviors by deterministic classification networks. In order to classify such ambiguous behavior sequences, it is e#ective to employ probabilistic sub-networks such as Hidden Markov Model with its learning methods. Consequently, by integrating those deterministic networks and probabilistic networks into the Bayesian Network, our system were able to successfully classify behaviors of rule violations, accidents, near miss situations distinguishing from ordinary behaviors in tra#c images.