A design of tracking system with multiple tracking methods deployed on Hisilicon platform
Ciyu Long, Weihua Wang, Chuwei Li · Seventh Symposium on Novel Photoelectronic Detection Technology and Applications · 2021
The research of visual tracking is of great significance for intelligent video surveillance, unmanned vehicle, manmachine interaction, intelligent robot, unmanned aerial vehicle reconnaissance and other fields. It has been a hot spot in the field of computer vision and pattern recognition. It provides a fundamental component for high-level video understanding applications such as motion analysis, event detection and activity recognition. The target tracking system mounted on uav platform is of high application value. However, in the actual development process, it is difficult to develop due to the fact that the Hisilicon platform has a series of special hardware modules and only supports the deep model under the framework of Caffe. We deployed some discriminative correlation filter algorithms and deep learning tracker on Hi3559A, and made a comparison of their performance and speed.