Real-time Target Tracking Based on PCANet-CSK Algorithm
Zhenghua Hu, Xianmei Wang, Kangming Xu, Pu Dong · 2019
This paper presents a real-time target tracking method by combining PCANet and CSK. To speed up PCANet feature extraction, we give a lightweight PCANet to simplify the structure of PCANet. In tracking step, in order to improve tracking performance, we integrate the scale adaptive module into traditional CSK algorithm and optimize its model updating mechanism inspired by LMCF. The experimental results on OTB50 validate the effeteness of our method. Compared with the traditional CSK algorithm using gray features, the tracking success rate of our method is about 26% higher, and the tracking accuracy increases about 29%.