WCSE 2017
ISBN: 978-981-11-3671-9 DOI: 10.18178/wcse.2017.06.008

Accelerating The JPDA Filter on CPU-GPU Platform

Wen Sun, Lin Gao, Xu Tang, Ping Wei

Abstract— This paper addresses the computational problem of Joint Probability Data Association (JPDA) filter in multitarget tracking with dense clutter. A parallelization implementation for JPDA on graphics processing unit (GPU) platform under the Compute Unified Device Architecture (CUDA) framework is presented. Where the JPDA filter is resolved into two parts in the view of computation. Specific parallelization scheme is designed for each part. Simulation results show that with the assistance of GPU, JPDA achieves up to a 55x speedup than its CPU implementation under dense clutter, while maintains the same computational accuracy to CPU method.

Index Terms— multitarget tracking, dense clutter, compute unified device architecture (CUDA), graphic processing unit (GPU), real-time implementation

Wen Sun, Lin Gao, Xu Tang, Ping Wei
Center for Cyber Security, School of Electronic Engineering University of Electronic Science and Technology of China, CHINA

[Download]


Cite: Wen Sun, Lin Gao, Xu Tang, Ping Wei, "Accelerating The JPDA Filter on CPU-GPU Platform," Proceedings of 2017 the 7th International Workshop on Computer Science and Engineering, pp. 48-52, Beijing, 25-27 June, 2017.