A Particle Filter Based Algorithm For The Tracking Of Dim Moving Point Target In IR Image Sequence

Extensive use of infrared imaging system makes infrared target which in image sequences tracking technology has a rapid development, most of time, the improvement of traditional method is the main research aspect. In recent years, for much universal applications, a target tracking technology based on particle filter become a new developmental field.Among the current domestic and foreign literature, target tracking research based on particle filter all rely on obvious characteristic , such as contour, texture, color and other information, which the dim moving point targets in this paper don't have. Under this challenge, considering characteristics of the dim moving weak point target in image sequence, this paper give systematic study on key technologies. The paper present a method in course of the tracking based on single frame detection technology which gets measurements from a single frame, effectively raising the real-time. A detailed analysis on related model, including system model,data probability model,noise model and so on. Facing the measurements data fusion and classification in multiple targets tracking, the paper adopt fuzzy clustering and data association techniques to fuse the single frame detection measurements data which are from different tracking windows. Under the condition of the target's initial information is already achieved successfully, the tracking technique proposed in this paper samples from target state, calculates the weights of the samples, and then uses the weighted particle to represent the target state, via the recursion iterativation of prediction and updating of each particle to get the moving state of each target at each discrete time. At the same time, solving the problem of lack of sample particles to some extent. Upon the completion of a single target real-time tracking, the paper had a research on multi-target tracking.Experimental analysis and results showed that the algorithm achieved real-time, dynamic stability and robustness while tracking the random moving targets in high clutter environment, having a certain application value.