Automatic Extraction Of Land Use Information Based On Remote Sensing Image

Remote sensing (RS) image classification is an important part of remote sensing study. How to improve the precision of RS image classification is a critical problem in the research of RS application.In recent years, with the great improvement of the algorithm of combination of multiple classifiers, the classification of multiple classifiers combination has become an important method for the application of RS image classification. This study first reviewed and analyzed the latest progress of research on the domestic and foreign RS image classification, especially the classification of multiple classifiers combination; based on this, the classification of land use of color infrared Navigation Remote Sensing image of the typical area of Huairen County of Shanxi province was investigated. Before the classification, the atmospheric correction, the independent component analysis, the minimum noise fraction, the texture filtering, and the image fusion were used in pre-processing the image; and then, considering the texture of surface features, ROI samples were trained based on texture fusion of the image. Besides, supervised classification was made to eight sub-classifiers, and furthermore, three of them were selected to be combination which were of more precision and better diversity in classification. The simple vote classification, maximum probability category method and fussy integral method were combined together according to certain rules. The experimental image was classified by using ENVI remote sensing software and Visual Programming Language IDL. According to the error matrix analysis, the overall precision of the classification of multiple classifiers combination was 93.53%, while the most precision of the single sub-classifier was only 81% of SVM. Comparing with SVM, the overall precision improved by approximately 12%,and the Kappa coefficient improved from 80% to 92.32%. Finally, post-classification was done to the result of classification, and the surface features of classification were exacted to be applied in our production and living.This result indicates that the classification of multiple classifiers combination is an effective classification method, furthermore, it can improve the classification precision of RS image. Comparing with the traditional methods, the multi-classifier classification method has a better extensibility, you can design a better algorithms of combination to improve the classification performance by multi-classifier combination methods.