Video and LIDAR Based Object Tracking Using Small-Region Growth

Constant fluctuations in outdoor environments cause many difficulties for object tracking. One of hard problems is to get high quality images. Indeed, with low quality images, many visual based methods may become invalid due to the missing of feature information required for object tracking. To address this sort of issue, laser range finder (LRF) measurements and raw pixel information based method is proposed in this paper. Given an object, it makes use of raw pixel information to locate interesting small region firstly. This region is then mapped into LRF space in which the region is grown to a complete region covering most part of the target through a bidirectional growth strategy. In the experiment, the proposed method is tested in a long-term video, and the results show its effectiveness.