Embedded vision tracking system based on DSP+FPGA

LU Guozhi1, SHI Jinghui2, PENG Dongliang1, GU Yu1

(1. Fundamental Science on Communication Information Transmission and Fusion Technology Laboratory, Hangzhou
Dianzi University, Hangzhou 310018, China; 2. Captial Spaceflight Machinery Company, Beijing 100076, China）

Abstract: Aiming at the problem of realtime requirement in visual surveillance applications, embedded vision tracking system based on DSP+FPGA architecture was proposed. Video capture and image preprocessing procedures, including autoexposure, color interpolation, median filtering, and white balance, were first carried out using FPGA as main controller. DSP processing results were then exported into host computer for display through PCIE interface. Highquality images were obtained from FPGA to DSP through EMIFA interface, and meanshiftbased tracking algorithm was adopted to track target in DSP. The design of module interface and implementation of main algorithms were introduced, and system optimal design was performed. The experimental results indicate that the system can obtain highquality images, where image brightness is moderate and color temperature is corrected. The target can be tracked stably and in time after code optimization. The average processing time per frame for tracking algorithm is 13 ms, which can meet realtime requirement.

Key words: FPGA; DSP; image preprocessing; visual tracking; meanshift