Summary
International Symposium on Nonlinear Theory and its Applications
2017
Session Number:C2L-D
Session:
Number:C2L-D-3
Model Free Object Tracking Using CNN Features and Color Information
Keiichiro Adachi, Kazuhiro Hotta,
pp.688-691
Publication Date:2017/12/4
Online ISSN:2188-5079
DOI:10.34385/proc.29.C2L-D-3
PDF download (502.8KB)
Summary:
Object tracking under occlusion is a challenging problem. In this paper, we propose a robust object tracking method under occlusion by adaptive integration of 2 kinds of trackers. Our tracking method is based on CNN features. However, the maximum value in the response map obtained by CNN features becomes small under occlusion. Thus, we can judge the occurrence of occlusion from the change of maximum value. When the target is occluded, we change the learning rate of a tracker using CNN features. A tracker using color information is also integrated because the tracker is robust to occlusion.