Implementation of a Multiple Frame Assignment Tracker in a CPU-GPU Integrated Environment

Abstract

<p>The Multi Frame Assignment (MFA) tracker solves the data association problem as a constrained optimization for fusing multiple sets of data to the tracks with an Interacting Multiple Model (IMM) estimator.</p> <p>With the rapid development of parallel computing hardware such as GPU (Graphics Processor Unit) in recent years, GPGPU (General-Purpose computation on GPU) has become an important topic in scientific research applications. However, GPU might well be seen more as a cooperator than a rival to CPU. Therefore, exploiting the power of CPU and GPU in solving the MFA tracker algorithm based on CPU-GPG integrated computing environment is the focus of this thesis.</p> <p>In this thesis, a parallel MFA algorithm implementation based on CPU-GPU integrated computing model to optimize performance is presented. The results show that the algorithm increases the average performance by 10 times compared with the traditional algorithm. Based on the results and current trends in parallel computing architecture. it is believed that efficient use of CPU-GPU integrated environment will become increasingly important to high-performance tracking applications.</p>