文章基本信息

标题：A Fast Contextual Classifier for Information Extraction from Remotely Sensed Imagery
本地全文：下载
作者：Peng Gong
期刊名称：ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences
印刷版ISSN：2194-9042
电子版ISSN：2194-9050
出版年度：1992
卷号：XXIX Part B7
页码：346-351
出版社：Copernicus Publications
摘要：A new contextual classifier has been developed and evaluated for information extraction from remotely sensed imagery. Thealgorithm is computationally very efficient, and experiment indicated that it can achieve more accurate results than theconventional maximum likelihood classifier and some commonly-used texture/contextual algorithms. The new contextualclassifier includes two basic procedures: grey-level vector reduction and frequency-based classification. In grey-level vectorreduction, the number of grey-level vectors in multispectral space is reduced using a new data-reduction algorithm throughrotating multispectral space into eigen space. As a result, the multispectral data are reduced to images of one feature dimensionwith the loss of relatively little information. Each grey-level vector-reduced image is then used in the frequency-based procedureto derive useful information. The frequency-based classification procedure includes a grey-level vector occurrence-frequencyextractor, a minimum-distance classifier and an accuracy evaluator. Landsat Thematic Mapper data have been used to illustratethe potential of the new algorithm. We have emphasized on land-use classification since land use is a cultural concept that isdifficult to be mapped directly using remote sensing data. The potential of using the grey-level vector reduction algorithm for fastclustering has been discussed as well