A Fuzzy Hidden Markov Model for Traffic Status Classification Based on Video Features

a novel approach for traffic status classification based on video features is presented. The proposed algorithm extracts DCT and Motion Vector features from traffic video sequence. And a Fuzzy C-Means Hidden Markov Model (FCM-HMM) is used to classify traffic status. Firstly, a method of extracting statistic features from Discrete Cosine Transform (DCT) and Moving Features is described. The extracted features described in this paper are good reflections of traffic status for the whole inspected area. They are also independent of camera setups and illumination conditions. Traffic status is classified into four states. And a Fuzzy C-Means Hidden Markov Model is trained to describe these traffic states. Then a dynamic decision process is used to give a most likely traffic state sequence after calculating video features. Experiment results prove that the proposed algorithm can not only give high-accuracy results but also can save much of memory resource and computing speed.