基于小目标孤立度的小目标检测算法

摘要

针对在红外图像序列中检测小目标的问题,本文提出了一种新型的单帧检测算法,该算法包括了两个部分:基于小目标孤立度的背景抑制和一种基于目标信杂比的自适应阈值分割.其中,小目标孤立度为本文提出的一种衡量像素属于待检测小目标的量度,与传统方法中把小目标视作图像中高频或者局部突出的思想不同,该量度着重考察目标的"孤立"的特性.实验结果显示,该算法的背景抑制效果明显,尤其对运用传统算法容易发生抑制不充分的建筑物边缘的有很好的抑制能力,与自适应阈值算法结合后的单帧目标检测也取得了满意的检测效果.%The problem of detecting small target in infrared video sequence was addressed. Instead of being regarded as prominence or high frequency part, the inherent isolated attribute of small target was taken into account and redefined as small isolated regions in image. Based on this idea, a new measurement called Small Target Isolation Degree (STID) was developed for evaluating the likelihood of a pixei as the center of a small target. By calculating STID in every pixel in image, the background clutter was suppressed. Then a novel adaptive threshold was proposed to extract the precise location of small target. Incorporated with the two methods, a single frame small target detection algorithm was built. Its high performance was then proved in a serial of experiments.