Modified pipeline algorithm for detection and tracking of infrared point targets

Abstract

A new method, namely the closed-loop pipeline, is proposed for the detection and tracking of point targets in heavy cloudy background. At first, a preprocessing step, i.e., a local contrast threshold is used to remove slowly changing clutter, then the "and logical" pipeline consisting of three frames are used to filter out the noise based on the highly consistency and continuity of the targets in temporal-spatial domain. After the two steps, the sequence is projected into one frame and the trajectory of the target is accumulated in the test pipeline. The target trajectory is detected by cluster analysis and smoothed by finding zero-crossing points; moreover the next searching window is identified by the achieved trajectory segment. So compared with the old open form pipeline, the new structure can predict the tracking window and it is advantageous over the traditional one in terms of searching space, computational complexity and clutter resistance. The experiments show that the proposed method can detect and track dim point targets with arbitrary trajectories accurately, and can also predict the searching windows efficiently.