Online Object Tracking Based on Depth Image with Sparse Coding

Shan-Chun Shen, Wei-Long Zheng, Bao-Liang Lu

Abstract: Online object tracking is a challenging problem because of changing environment including diverse illumination and occlusion conditions. The emergence of commercial real-time depth cameras like Kinect make online RGBD-based object tracking algorithm become a focus of research. In this paper, we propose a robust online depth image-based object tracking method with sparse coding. We introduce sigmoid normalization for local depth patch. In order to recovery from tracking failure in condition of heavily occlusion. we present a detection module based on PCA bases. Experiments show that our method exceeds original color image-based method in case of environment changes.