Small target fusion detection algorithm via image neighborhood entropy and univalue segment assimilating nucleus principle

摘要

Small and dim targets detection in the presence of strong background clutter is a challenging problem faced in many applications including space surveillance and missile tracking. To solve this problem, a new fusion detection algorithm applied image neighborhood entropy and univalue segment assimilating nucleus (USAN) principle is presented. In this method the neighborhood entropy is used to locate small and dim targets. And the USAN principle is used to extract some geometry features of targets including edges and inflexions. Based on these results, image fusion method is used to detect real targets from noise and false targets. Finally, an iterative image threshold technique is proposed to label and locate targets more precisely. Simulations and experiments show that the new fusion detection algorithm takes advantage of the USAN principle and the neighborhood entropy method and it can detect small dim targets robustly, fast and efficiently.