FPGA-based Real-Time Object Tracking Using a Particle Filter with Stream Architecture

This paper deals with the real-time FPGA implementation of the posterior system state estimation in dynamic state-space models using a particle filter. The system is constructed by parallel resampling (FO-resampling) algorithm on a stream-based architecture. In particular, the system consists of three steps: prediction, likelihood calculation and resampling. Since the resampling is accomplished in a synchronized area, our approach enhances the object tracking system especially efficiency and performance. The result shows that the amount of FPGA resource utilizes for the simulation of red-color soccer ball tracking compared with the available usage. Moreover, we evaluate the tracker detection rate and the accuracy of object tracking with the calculation of average and maximum tracking errors.