Tracking In Aerial Video Sequences Based On SIFT Compensation And Particle Filter

Aerial video target tracking is a research hotspot of target tracking field. It has been widely used in many military and civilian areas, such as UVA, image guided missiles, aerial disaster analysis, transportation monitoring based on air platform, and so on. The difficulty of aerial target tracking is that vehicle movement brings about interference to tracking. With the analysis of the characteristics and the difficulties of aerial video, the paper puts forward a kind of aerial video target tracking solutions.This paper aims at the target tracking algorithm based on Bayesian in aerial vehicle motion problems, and adopts SIFT feature points image registration for vehicle motion compensation. A solution based on particle filter and SIFT compensation for aerial video target tracking is presented. With the assumption that both object and camera move at the same rate between the frames in the short term after SIFT matching algorithm, the middle frame's object position is sent to particle filter, and location and diffusion of the particles are modified. The presented algorithm makes full use of the time information between the frames to modify the frame subtraction. Comparing the presented algorithm with the particle filter with Euclidean distance weighted and the particle filter with Bhattacharyya distance weighted respectively, the experiment results show that it is not sensitive to target shape, and has better positioning accuracy and robustness.