针对目前小目标检测存在的难点, 提出了一种基于时空域融合的小目标检测方法。该方法首先运用形态学 Top-hat进行空域滤波, 将原图像中的大部分背景及杂波抑制, 时域上采用三帧差分方法增强目标。融合后对图像进行自适应阈值分割, 得到潜在目标点, 最后根据目标运动的连续性与规则性采用邻域判决法滤除虚警点, 检测出目标的运动轨迹。仿真结果显示该方法能较好检测出复杂背景下低信噪比运动小目标。

For the problem in small target detection, a temporal-spatial fusion algorithm is proposed. This algorithm first suppresses background by morphologic top-hat transform filtering in spatial domain, then uses three sequential frames to enhance target. It uses adaptive threshold to segment the frame and gain the potential targets, finally detects the target and excludes the false alarm points based on the continuity and regularity of target’s movement. The experimental results show that small target with low SNR can be detected in complex background with this algorithm.