基于模糊分类的弱小目标检测方法

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。本文直接从待分类图像入手提取出不同的类别区域，这样得到的分类模板就准确的体现了当前图像的不同类别，在此基础上进行分类就能得到图像的准确的类别分类从而实现弱小目标检测。首先，对红外天空背景弱小目标图像进行分析，将图像中的三类物体：净空、云及弱小目标细分为11个类别区域；其次，定义了类别特征矢量并基于此提出了类别核的定义；再次，根据类别核的定义从待检测图像中提取出11类区域的类别核；最后，根据模糊分类的理论，定义了类别相似系数和类别贴近度，通过类别核对图像进行分类和类别归并，保留弱小目标类别完成检测。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分类，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。

Approach to dim and small target detection based on fuzzy classification

In order to achieve robust dim and small target detection in the infrared cloud clutter, a new approach based on fuzzy classification is proposed. Different kinds of class regions are extracted from the query image to get several classification models, which can describe different classes in the image exactly. Classification based on such models will classify the image effectively and achieve robust dim and small target detection. Firstly, analysis of dim and small infrared target image is performed. Eleven kinds of class regions are proposed to describe sky, cloud and the target in the image. Then class feature vector and class kernel are defined. Class kernels of eleven class regions are extracted from the query image. At last, class similar coefficient and class similarity degree are defined according to the fuzzy classification theory. Image classification and class merge are performed. Target detection is achieved by reserving dim and small target class. Experimental results show that the proposed method describes different kinds of regions in the dim and small infrared target image effectively and provides robust dim and small infrared target detection in heavy background clutter.