基于多光谱与显著性的红外弱小目标融合检测

基于多光谱与显著性,提出一种红外弱小目标融合检测算法。该算法旨在将从多光谱探测器获得的同一场景的多光谱图像信息组合到一起,利用它们在时空上的相关性及信息上的互补性,提高系统的检测性能。采用一种新的基于人类视觉注意机制的显著性方法,该方法能够快速准确找到图像中的显著性区域;将目标看作一类,背景和干扰点看作另一类,选取离差平方和准则,使类内距离最小,类间距离最大;训练出融合参数,得到融合后的显著性图像。通过设定的门限值二值化,可以看出该融合方法能很好地将目标与背景分开,从而检测出目标。

Fusion and Detection of Infrared Dim and Small Targets Based on Multi-Spectral and Saliency

A new fusion and detection algorithm for infrared dim and small targets is proposed based on multi-spectral and saliency.The algorithm is the combination of information acquired from multi-spec-tral sensors in the same scene by using temporal and spatial correlation and their complementarity, aims at improving the detection performance of system.Based on human visual system, saliency is adopted, which can let computer find the interesting regions fast and correctly.The targets are considered as a class, background and interference points are another class.This algorithm chooses sum of squares of de-viations as a criterion to make the minimum distance within the class and the maximum distance between the classes.Parameters of fusion are trained and the saliency image is obtained.Experiments show that this algorithm can well separates the object and background, thereby detecting the targets.