Research On Tracking Algorithm Of Point Targets In Infrared Image Sequence Based On Gaussian Particle Filtering

Infrared Search and Track(IRST)system provides detection and tracking of potential targets by receiving infrared radiation and it is equal to a passive radar.It has passive detection and invisibility, high resolving ability, fine anti-interference, simple system structure, small size, lightweight, low power expenditure and high reliability. As typical airborne detecting devices equipped in fighters, they are attached much importance. How to track maneuvering target and multiple targets using IRST systems is a significant research. In recent years, the theory and method of target tracking have been greatly developed and become one of the hot areas of research and engineering praceice.This paper presents the development and classification of IRST systems detailedly, expatiates the origin and development of target tarcking,and recounts the classification and realization principle briefly, analyzes and compares many nonlinear filterings, focuses on particle filter and the Gaussian particle filter. To solve infrared maneuvering target tracking problem, two new tracking algorithms are designed,which are Current Statistical Model Tracking Algorithm based on Gaussian Particle Filtering and Interacting Multiple Model Tracking Algorithm based on Gaussian Particle Filtering. The simulation shows that the robust and tracking accuracy of the two tracking algorithm are superior to those of the others. To solve data association problem in multiple targets tracking, we analyze some typical data association algorithms including the Nearest Neighbor algorithm(NN),Probabilistic Data Association Filtering(PDAF),Joint Probabilistic Data Association Filtering(JPDAF) and Multiple Hypothesis Tracking(MHT). With the Nearest Neighbor algorithm and Current Statistical Model Tracking Algorithm based on Gaussian Particle Filtering, a new fast data association algorithm is designed. The simulation results conduct to verify the effectiveness of the proposed method.