IR Small Moving Target Real-Time Tracking Based On Multi-level Filter

Since infrared imaging system deals with 3D space-time signal and is a kind of passive detecting system, comparing to other tracking systems, it has a lot of features, such as concealment, anti-jamming and high precision tracking performance. With the developing requirements in applications, the targets are required to be detected as soon as possible and as far as possible. So there is much interest in studying of the detection and tracking technology of dim small moving target from image sequence.In this dissertation, we mainly study the technology of real time small moving targets tracking in IR image sequences. The main contribution of the dissertation is as follows:(1) The technique of background suppression in IR image is discussed and studied principally. Based on the fruits developed, a modified structure based on primal multi-level filter is proposed. Compared to other preprocessing technologies, experiments demonstrated the approach based on fusion of outputs from modified filter is effective in detecting multi-scale small targets in clutter.(2) On the basis of IR imaging tracking system, mainly studies targets detection and tracking techniques, including the classical methods for IR small target tracking in high clutter background and multi-targets tracking(MTT). Compared to some familiar methods of data association, a data association algorithm based on multiple features of target is realized.(3) By the different regional features of small targets and strong cloud, A method for small target detection in roughly cloud is proposed. Based on target detection, a method of tracking small moving targets is discussed on IR image sequences. At last, an algorithm based on fusion two IR bands is realized, by which the real target can be tracked steadily while IR bait is appearing from image sequences.(4) A real time IR tracking system based on multi DSP of TMS320C6414 is discussed. First, the fast calculating method of multi-level filter is proposed. Besides, program optimizing of algorithm realizing depended on C6000 is also discussed. At last, a parallel structure based on master-slave hardware system is adopted, and the problem of algorithm distribution is discussed.