基于分类的红外云层背景弱小目标检测方法

A New Approach of Small and Dim Target Detection in Cloud Cluster Infrared Image Based on Classification

提出了一种新的基于模糊分类的红外云层背景弱小目标检测方法。根据红外成像的特点,将红外云层背景弱小目标图像分为三类:边缘类、净空及云中类、弱小目标类;对不同类别图像进行分析,建立了分类模型,并定义了方向特征矢量,将其作为类别的特征矢量;根据模糊分类的理论,定义了类相似系数来判别图像中每一个像素的类别属性,保留弱小目标类的像素点完成检测。实验结果表明,该方法能够对红外弱小目标图像中不同类型的区域进行准确的分类,从而较好的实现了对低信杂比的复杂云层背景图像中的弱小目标检测。

A new method is proposed to detect infrared small and dim targets in cloud cluster image based on fuzzy classification. Firstly, according to the infrared imaging principles, dim and small targets in cloud cluster image can be divided into three classes, which are the edge class, the inner cloud and clear sky class, the target class. By analyzing different classes, the classified models for each class are established. And then four-direction pixel feature vector is defined to describe different classes. Finally, based on the fuzzy classification theory, class correlation coefficient is defined to classify each pixel of the infrared image. Experimental results show that the proposed method can separate dim and small targets from the cluster in low signal-to-clutter ratio （SCR） infrared cloud images efficiently.