Target detection of maritime search and rescue: Saliency accumulation method

Target detection using visual image sequence is an important task of maritime search and rescue. A novel saliency accumulation approach is proposed. Each frame of image sequence is firstly transformed to LAB space. The saliency maps for two color channels and intensity channel are obtained by spectral residual in frequency domain and the local saliency in space domain taking advantages of both approaches. Then saliency map of every frame is fusion of those from each weighted feature. In order to reduce the sea clutters or detect the weak small target, the saliency maps from successive frames are accumulated to reach a binary saliency map by applying a threshold. The approach is simple and effective. Experiments are conducted to demonstrate the validity of proposed approach for target detection in maritime search and rescue.