摘要： 提出使用特征提取的方法来识别遥感图像中的小目标。介绍了不变矩法和高阶累积量法两种用于目标识别方法的原理和如何构造它们的不变特征集的过程。在论述了这两种不变特征集的构造方法的基础上，对遥感图像采用多元识别方法。实验结果证明：经过图像预处理后，采用不变矩法和高阶累积量法进行多元搜索有很高的准确度。

关键词: 不变矩法, 高阶累积量法, 特征提取, 多元识别

Abstract: This paper puts forward a method of feature extraction to recognize small target in remote sensing images, introduces the theory of two kinds of target recognition algorithms: moment invariants algorithm and high order statistics algorithm, and how to construct their invariable feature subset. Based on the feature subset of these two algorithms ,it uses multiplex recognition method on some remote sensing images. The result shows good accuracy.

中图分类号: 

• TP751