Dim targets detection and tracking by self-adaptive segmentation and particle filter in starry images

An effective method for dim and small multi-targets detection and tracking through successive CCD images in complex starry background is put forward in this paper. Optical starry background images contain a lot of interference noise besides the moving targets. Firstly, self-adaptive threshold segmentation can play an important role in eliminating noise and improving detection rate. Furthermore, back neighborhood frame correlation (BNFC) is proposed to detect and locate the target, which is sheltered by bigger interfered stars. After detection framework acquiring the location of moving targets, particle filter which has nonlinear filtering feature is applied to track the trajectories for multi-targets in real-time. Experimental results show that by using the adaptive target detection and improved particle filter, the trajectories could be achieved at a relative low signal to noise ratio (SNR ≥ 3.5) in the case of multi-targets detection and tracking in real time. The method has good prospect for engineering application.