Detection Algorithms For Infrared Dim Small Target And Performance Analysis

Low observable small target detection, recognition and tracking algorithms are key technologies to extend the acceptance distance of guidance system and enhance the capacity of defense system. Based on the assumption the target is point-like and slow moving in the outer space and cloud clutter, the detection algorithms are particularly studied in this dissertation. .In chapter 2, the two assumptions, one is that the background intensity is Gaussian distribution; the other is that the trajectory of moving point target is approximate to a straight line, are shown. Then author presents a track-association algorithm based on the least square predictor besides analyzing the detecting performance of weight dynamic programming (DP). This chapter emphasizes contents on the analysis of the algorithm in theory and thoroughly dissertate its theory and structure. Finally it shows with the simulation experiment results that the presented algorithm has a lot of good qualities, which are excellent performance, simple structure, easily realized with parallel hardware. The researched algorithm is well done in real work. In chapter 3, cloud clutter, noise and target pixel temporal profiles in image sequences are well studied. On account of spatial correlativity of cloud clutter, the author adopts a pre-processed clutter background canceling method with max-mean filters. Due to the singularity of target arrival position in the temporal profile, the detection algorithm based on recursive variance filter is expatiated in this chapter. From the result of experiments, the author made the conclusion that the algorithm is not only detecting point target under cloud clutter very well.At last, the author summarizes achievements in the thesis and points out the future work.