More Robust Object Tracking via Semi-supervised Metric Learning and Edge Detection
Liujun Zhao, Weizheng Kong, Ming Qin · 2018
In this paper, we propose a novel object tracking method via semi-supervised metric learning and edge detection. Firstly, we construct the object appearance model by extracting sparse codes features on different layers to exploit local information and holistic information. To utilize unlabeled samples information, the semi-supervised metric learning is introduced to measure candidates. In addition, based on ranking results from the learned metric classifier, we adopt an edge detection to alleviate the error accumulation. Both the proposed tracker and several selected trackers are tested on some challenging videos, where the target objects undergo pose change, illumination and occlusion. The experimental results demonstrate that the better performance of the proposed tracker compared with state-of-the-art methods.