Tracking with Stereo-vision System for Low Speed Following Applications

Research in adaptative cruise control (ACC) is currently one of the most important topics in the field of intelligent transportation systems. The main challenge is to perceive the environment, especially at low speed. In this paper, we present a novel approach to track the 3D trajectory and speed of the obstacles and the surrounding vehicles through a stereo-vision system. This tracking method extends the classical patch-based Lucas-Kanade algorithm by integrating the geometric constraints of the stereo system into the motion model: the epipolar constraint, which enforces the tracked patches to remain on the epipolar lines, and the magnification constraint, which links the disparity of the tracked patches to the apparent size of these patches. We report experimental results on simulated and real data showing the improvement in accuracy and robustness of our algorithm compared to the classical Lucas-Kanade tracker.