Independently moving object detection and tracking using stereo vision

We present an approach for detecting and tracking independently moving objects from a mobile platform using uncalibrated stereo cameras. Firstly, scale invariant feature transform (SIFT) features are detected and a novel multiview matching method is proposed for simultaneous stereo matching and motion tracking of the detected features. A multiview geometric constraint, derived from the relative camera positions in pairs of consecutive stereo views, is then derived for independent motion detection. Finally, a dimensional variable particle filter is introduced for joint detection and tracking of multiple independently moving objects. Experimental results on real-world stereo sequences demonstrate the effectiveness and robustness of our method.