Abstract
In this paper, we propose a fast object tracking method used in real-time application. The tracking method is based on compact kernelized correlation filters. Traditional, gray values are used to describe target feature. In order to achieve relatively fast performance, we adopt Principal Component Analysis (PCA) to simplify Histogram of Gradient (HOG) as feature descriptor, which can significantly reduce computation resources. Our method is implemented on Texas Instruments (TI) multi-core digital signal processor (DSP) TMS320 C6657, which can achieve 60 frame per-second (fps). Finally, we conduct both quantitative and qualitative experiment to show the robustness and real-time performance of our method.
Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
