Study on image real-time interpretation based on particle filter

In order to satisfy to the real-time requirement of image interpretation system in photoelectric measurement equipments, a kind of hardware acceleration system with MIMD distributed multi-processor architecture based on SOPC technology is designed. The particle filter algorithm is proposed to process image interpretation for state estimation problem of nonlinear and non-Gaussian system. This algorithm does not involve conventional linearization transform, and has approximated the posterior probability density by a set of discrete particles. Therefore the approximate optimum result is educed. It has a high accuracy and a rapid convergence. Experimental results show that the algorithm be adequate to real time, accuracy and robustness, meets the requirement of image interpretation and possesses practical significance for engineering applications.