强阳光反射背景下红外舰船目标自适应分割

Adaptive Infrared Warship Target Segmentation in Strong Sun-glint Environment

针对海面有强阳光反射背景下现有红外舰船目标分割算法效果较差,提出了一种有效的红外舰船目标分割方法。利用均值漂移方法的不连续保持性滤波特性,抑制海面小面积强杂波干扰,同时又不损失舰船目标信息。对滤波后图像做行垂直方向梯度均值投影,获取舰船目标感兴趣区域。最后运用改进的行列均值自适应阈值分割算法,结合使用红外舰船目标的几何特征量达到自动分割的目的。

The traditional segmentation algorithm for infrared warship target effects poorly in dense or strong sun-glint sea-based environment. This paper presents a novel adaptive threshold segmentation algorithm to solve the problem. At first we filter an original image using mean shift technique and considering it’s discontinuity-preserving smoothing property so that we can cluster the small-sized strong sun-glint away from the warship target. Then we get the ROI by projecting the average vertical gradient of each row. Finally, we achieve the goal of automatic ship segmentation by the modified adaptive threshold segmentation algorithm based on the mean of each row and column. Furthermore, we extract several regional geometry characteristics to further suppress sun-glint interference.