Adaptive visual target detection and tracking using incremental appearance learning

Multiple visual target tracking is a challenging problem due to various uncertainties including noise, clutter, miss-detection and occlusion. In this paper, we propose an adaptive keypoint-based appearance model to represent the appearance of visual targets independent of their shape or type. We also develop an incremental learning algorithm to learn the appearance of targets in time. The proposed method utilizes a simple background subtraction method to prune insignificant keypoints and to adapt the target appearances in different frames. The experimental results presented on several video datasets show the effectiveness of our proposed method.