A Small Target Detection Method based on Human Visual System and Confidence Measurement

Robust small target detection of low signal-to-noise ratio (SNR) is very important in infrared search and track applications.Due to the complex background, current algorithms have some unsolved issues about small target detection capabilities, such as detection rate, false alarm rate and speed. They generally improve one or two of the aforementioned detection capabilities while sacrificing the others. In order to pursue good performance in detection rate, false alarm rate and speed simultaneously, an infrared small target detection algorithm based on mechanism of visual attention and two layers structure was proposed.Firstly, we detected salient regions that may contain targets by morphological Top-hat and threshold operation. Then target recognition was performed by confidence measurement based on histogram in the salient regions. Experimental results show the proposed algorithm has ideal robustness and efficiency for real infrared small target detection applications.