The new approach for infrared target tracking based on the particle filter algorithm

摘要: Target tracking on the complex background in infrared image sequence is hot research field. It provides important basis some fields such as video monitoring, precision, and compression human-computer interaction. As a typical algorithms target tracking framework based filtering and data connection, particle filter with non-parameter estimation characteristic have ability to deal nonlinear non-Gaussian problems so it were widely used. There are various forms of density algorithm make valid when occlusion occurred or recover back from failure track procedure, but in order capture change state space, need certain amount particles ensure samples enough, this number will increase accompany dimension increase exponentially, led increased calculation presented. In paper particle filter Mean shift be combined. Aiming at deficiencies classic mean shift Tracking easily trapped into local minima Unable get global optimal under the background. From these two perspectives that "adaptive multiple information fusion" and "with framework combining", we expand Shift .Based previous perspective, proposed an improved infrared target fusion. analysis infrared characteristics basis, Algorithm firstly extracted gray edge character and Proposed guide above characteristics by moving thus we can new sports grayscale motion border feature. Then proposes a adaptive fusion mechanism, used integrate the Mean framework. Finally designed kind automatic model updating strategy further improve performance. Experimental results show algorithm can compensate shortcoming has too much computation, can effectively overcome fault mean easy fall extreme value instead global maximum .Last because information, approach also inhibit interference background, ultimately stability real-time the target track.