多种特征信息在高分辨率遥感影像目标识别中的应用研究

以Quickbird影像为研究对象,探讨了利用多种特征信息识别地物目标的技术方法.首先采用区域生长法将影像分割为若干个具有语义信息的对象,然后在此基础上提取对象的光谱、形状和纹理特征并进行描述,最后根据提取的特征参数,采用最近邻方法将影像分为建筑物、公路、铁路、水塘、耕地、林地和荒地7类地物目标,综合分类精度达到91.03%.研究表明,多种特征信息的综合利用,在目标分类与识别方面明显优于传统的基于单一光谱特征的方法,在一定程度内提升了遥感信息的智能化水平.

Recognizing and Classifying Object by Using Multiple Features

Take Quickbird image as example,this paper discusses the technique of target recognition using multiple features information.Firstly,the image was segmented into semantic segments by Region-Growing method,and then the spectral,shape and texture features was extracted and described,based on these feature parameters,the image was classified to seven classes,they are buildings,roads,railway,lake,vegetation,forest and wasteland,and the total accuracy is 91.03%.The results indicate that the technique integrating multiple feature information overcomes the limitation of traditional pixel-based approach,besides,it improves the intelligence of information extraction to some degree.