Author(s)

Rui Li and Jirong Lian

Corresponding Author

Jirong Lian

Abstract

Aiming at the problem of real-time high-speed visual tracking and how to improve the success rate and accuracy of visual tracking without significantly increasing the computing performance requirements is the key to solving such problems. Also aiming at the problem that historical information of traditional visual tracking algorithms is easy to lose, this paper proposes an algorithm that uses tree structure and multi-convolutional neural network to jointly estimate the target state and dynamically updates the model. After experiments, the performance and efficiency of the algorithm are higher than traditional algorithms, and it can help to deal with high-speed visual tracking problems.

Keywords

Convolutional neural network; Visual tracking; Deep learning; Machine vision