Detection of dim point target with low contrast

An image sequence-based method is proposed for target detection within a large-field bright background from a long distance. Optimal estimation of image background is obtained using adaptive prediction of local background. For the image with background subtracted, energy accumulation and median filter are used to further remove clutter and to improve signal-to-noise ratio (SNR). Then the targets are located by double window filtering with buffer window which further segments the targets from background. With improved high order correlation, the true dim target is eventually detected with a small amount of calculations. The proposed method is implemented in field experiments. Using three order correlation, its detection probability achieves 98% when the local SNR is higher than 0.3.