The Classification Technology Research Based On Hyperspectral Data

During the recent two decades, hyperspectral remote sensing has been playing an important role in both military and civil applications. It's urgent to develop fast and accurate methods in order to discover the interested information from the huge data which were produced by hyperspectral sensors. Based on the properties of hyperspectral image data, effective classification algorithms for extracting most information of groundcover types from hyperspectral image data are studied in this thesis.The classification and recognition of hyperspectral image have been set as the aim of this study , introduced an extraction method of endmembers which based on the hyperspectral and the mixed-pixel decomposition model. Proposed a classification method with the combination of the traditional classification method and the pixel space characteristic. Finally, get the conclusion with the comparison of tradition high spectral classification technology and union pixel space characteristic classification technology. To sum up, this paper mainly has mainly done the research on the following aspects:1),Introduced the Pixel Purity Index method, the Convex Cone Analysis method and Endmember Extraction Algorithm based on RMS error analysis method. At the same time, made an experimental comparison of the Pixel Purity Index method and the Convex Cone Analysis method by using the existing hyperspectral data。2),Analysed several modeling methods of Decomposition of mixed pixels in detail, namely the linear spectrum mixed model, the non-linear spectrum mixed model and the fuzzy anatomic model.3),Using the extracted endmember and to withdraws the endmember to carry on certain solution to mix in the situation , introduced several traditional hyperspectral data classification methods, namely Maximum Likelihood classification, Artificial Neuron Network classification technology and Spectrum Angle Mapper.4),Studies in the predecessor in foundation , proposed the hyperspectral classification method of union pixel space characteristic, and has carried on the experimental comparison and analysis on this method and the traditional hyperspectral classification method. Made a conlusion that it is fully certain to enhancement classification precision by using the classification method of pixel spatial characteristic...