Based on the features of glint and clutter sea-sky blending background as well as point target with one pixel, a fast location algorithm for sea-sky blending area and an extended median-mean filter for all direction ring-neighborhood difference data based on pattern classification were designed as pre-processing methods. The all-direction difference was re-carried out again so that candidates of point target can be detected by segmentation with an adaptive threshold based on histogram. The morphologic filter was used to keep point targets with only one pixel. Finally the moving point target and its trajectory can be detected in sequence images based on trajectory matching and direct search if mismatching occurs. The simulation results with 220 frame images which were made by adding simulation point target in real images show that moving point targets with signal to noise ratio (SNR) not lower than 0.5 can be detected effectively: the correct detection rate is 91% in one frame and 95% in a sequence of three frames, in addition the false alarm in a sequence of three frames does not occur for the 22 frame images without simulation point target.
Zhu Fengyun, Qin Shiyin.Novel method for moving point target detection under glint and clutter background of sea-sky blending[J] JOURNAL OF BEIJING UNIVERSITY OF AERONAUTICS AND A, 2008,V34(12): 1384-1387