The content-based approach is utilized to effectively and efficiently extract lithology information from large scale remote sensing imagery.The framework of the proposed algorithm is described as follows:firstly,the large scale remote sensing image is decomposed into equal-size subimages.The spectral or textural features are then extracted from the subimages.Secondly,the user selects query subimages,which is able to represent each lithology characteristics within the image.Finally,the algorithm calculates the dissimilarities between query subimages qi and every subimage Ij.If the dissimilarity between a subimage Ijand a query subimage qi is the lowest,Ij is classified into the lithology class represented by the query subimage qi.Since the user is involved in the selection of the query subimages,the algorithm is able to effectively utilize the knowledge and experiences of the user to extract lithology information from the remote sensing images.Moreover,the computational complexity of the algorithm is relative low,and thus its extraction speed is faster than supervised or unsupervised classification algorithm especially on the large remote sensing imagery.%利用基于内容的方法快速有效的对大尺寸遥感图像的岩性信息进行分类提取.基于内容的方法的框架如下:首先将大尺寸遥感图像切割成等尺寸的子图,然后从子图中提取颜色纹理等特征,用户再从切割的子图中一一选出最能代表每一类岩性特点的查询子图qi,最后算法将用户选出的查询子图面与所有子图进行相异度计算,如果莱一个子图Ij与查询子图qi的相异度最低,筑将子图Ij归类为查询子图qi所代表的岩性.由于有用户参与选取查询子图,该算法能够有效利用用户的知识和经验对遥感图像的岩性进行分类提取.由于该算法计算复杂度较低,对大尺寸遥感影像的提取速度要快于无监督和监督分类算法.
展开▼
