Siamese network tracking algorithm based on reinforcement learning

Qingbo Ji; Yufei Qi; Hang Liu; Changbo Hou; Jianyang Gong

doi:10.1117/12.2671078

28 April 2023 Siamese network tracking algorithm based on reinforcement learning

Qingbo Ji, Yufei Qi, Hang Liu, Changbo Hou, Jianyang Gong

Proceedings Volume 12610, Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022); 126105M (2023) https://doi.org/10.1117/12.2671078
Event: Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 2022, Wuhan, China

Abstract

In the field of computer vision, object tracking is a hot research topic. At present, the coordinate box of most algorithms is vertical box. However, in practical application, the frame selection area of the object is expanded due to the rotation of the object, which makes it impossible to obtain the real motion angle of the object. To solve this problem, this paper proposes a Siamese network tracking algorithm (SiamRotate) based on rotation box reinforcement learning. In order to predict the rotation Angle of the object, firstly, we process the VOT2013 and VOT2014 datasets to increase the angle richness and used the processed datasets to train the angle decision-making network through the combination of supervised training and reinforcement learning. Based on the SiamRPN network, the tracking network obtains the feature response through the deep cross-correlation operation of the feature map of the two branches through SiamRPN, and then processes the feature response map through the angle decision network to predict the object rotation angle. The performance indexes of the proposed method reached 0.62(A), 0.21(R), 0.425(EAO) on VOT2016, and 0.61(A), 0.27(R), 0.399(EAO) on VOT2018, and the tracking speed reaches 138fps. Finally, comparative experiments show that the algorithm has good performance in speed and accuracy.

Citation Download Citation

Qingbo Ji, Yufei Qi, Hang Liu, Changbo Hou, and Jianyang Gong "Siamese network tracking algorithm based on reinforcement learning", Proc. SPIE 12610, Third International Conference on Artificial Intelligence and Computer Engineering (ICAICE 2022), 126105M (28 April 2023); https://doi.org/10.1117/12.2671078

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available