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：2008
卷号：XXXVII-B8
页码：999-1006
出版社：Copernicus Publications
摘要：Availability of new generation of hyperspectral sensors such as the Hyperion has lead to new challenges in the area of crop type mapping and agricultural management. Many crops like wheat and barley are spectrally similar and may not be discriminated by the normally available multispectral data. Although the existing hyperspectral data provide the possibilities for discrimination of crop types, but consideration of spatial variability between the adjacent pixels known as the texture data, can lead to more accurate results in a classification process. In this research, Hyperion data of an agricultural area located in south of Tehran, has been examined for discrimination of wheat and barley fields. The output bands of linear unmixing algorithm have been used as inputs for texture feature generation by different methods including the First Order Statistics of the Gray Level Co-occurrence Matrix, Geostatistics and Fourier Transform. Maximum likelihood classifier has been applied to classify the different combinations of linear unmixing outputs and texture features. Overall accuracies as well as the producer accuracies have been used as the evaluation criteria for different classifications. Results of this work have shown that the use of texture features generated from the output bands of linear unmixing algorithm lead to higher accuracies. Overall accuracy improved up to 7% and better discrimination between similar classes where obtained
关键词：Hyperspectral; Texture Quantization; Classification; Hyperion; Linear Spectral Unmixing
