Feature-based vehicle tracking using trajectory matching

In this paper, a new feature-based vehicle tracking system is proposed using trajectory matching for intelligent traffic surveillance. The proposed system extracts the corner features that are seen as vehicle and tracks the features using linear Kalman filtering. Then, the features that come from the same vehicle are grouped together. A new grouping algorithm using trajectory matching is proposed to make our tracking system robust enough to segment different vehicles of congested traffic or noisy objects. The simulation results show that the proposed system has demonstrated good performance on crossway traffic sequences. Copyright (C) 2001 IFAC.