IMM/EKF filter based classification of real-time freeway video traffic without learning

ABSTRACT

This paper addresses the problem of traffic variable estimation and traffic state classification of highway traffic, from video. To solve this problem, we propose to use the Interactive Multiple Model (IMM) filter with a multi-class macroscopic model. This filter runs two Extended Kalman Filters (EKF) to smooth the measured traffic parameters. In addition, the models’ probabilities that it provides are exploited to simply classify the traffic state as either free or congested, without the need for a training phase. The evaluation of the proposed system using simulated traffic parameters shows that it achieves a very accurate traffic state classification. The system was also tested in the real world, using video data acquired on a freeway by camera sensors. The obtained classification rates are comparable to those obtained by SVM classification, but at a significantly lower computational load.

KEYWORDS:

• Highway video traffic classification
• traffic estimation
• extended Kalman filter
• interactive multiple model
• learning

Disclosure statement

No potential conflict of interest was reported by the authors.