Real-Time Implementation of Particle-PHD Filter Based on GPU

This paper addresses the computational problem of particle-Probability Hypothesis Density filter (P-PHDF) in multitarget tracking. A parallelization implementation scheme for P-PHDF on graphics processing unit (GPU) under the Compute Unified Device Architecture (CUDA) framework is proposed. Simulation results show that nearly 20× speedup was achieved on GPU compared to its CPU version.