The Study Of The Method About Vegetation Information Extract Based On Independent Component Analysis

With data sources of remote sensing become more and more abundance and the methods of vegetation information extract are improved on too, from using spectral information only to classing based on combining the spectral information and the space information. But as there are a lot correlations between remote sensing data, the methods of vegetation information extract before didn't clear up these affections of correlation, and based on the low layer of the data merely, all these factors affect the assoeted precision.The independent component analysis is an effective method of data decomposition. The basic problems of signal processing are looking for a felicitous linear express,data compression and data yawp removing, and independent component analysis had been used to settled these problem successfully. The study of the method about vegetation information extract based on independent component analysis, utilize the method of independent component analysis to extract spectral information of the image, which using the high layer of spectral information better and eliminate the correlation effectively. By comparing the results obtained, the superiority of this method is proved.The theory of vegetation information extract, independent component analysis, grey scale co-occurrence measures and K-means dynamic clustering arithmetic are introduced first. Then the ASTER data as the data source of the region interest is showing. In order to eliminate the effect of atmosphere, we should do the atmosphere correlation to the original data, then use the method of Fast ICA to extract the independent component. And using this method we can clear up the correlation of every component well and effectively reduce the number of the dimension. In the last the independent component and texture character are combining to class using the k mean dynamic clustering arithmetic. In the result of classification the nine kinds of object are distinguished well, and the general niety of grading is 93.5369% with its Kappa coefficient reaches 0.9212. Then the result is compared with the results that gained by dynamic clustering combining the texture character and principal component and supervised classification using the spectral information only. In the last the method is validated superior.The method of independent component analysis is applied to extract the vegetation information and compared with the method based on the traditional principal component analysis. The results show that the method of independent component analysis based on high layer statistic can fully use the spectral information of the remote sensing image and effectively eliminate the correlation between every component. The dimensions of data treated with by independent component analysis are depressed greatly and the niety of grading is higher than treated with by principal component analysis. All this show that the method of independent component analysis makes for right disassemble of the data.Besides, in order to gain the reflectivity data of the ASTER data, FLAASH model of ENVI software is utilized to correct atmosphere, and the reflectivity data can improve the quality of the spectral information; Variance, second moment, entropy, correlation and contrast can represent the texture of image well, especially when the spectral information of the object is close to each other.With the development of the theory about independent component analysis, the abundance and improvement of the algorithm model and the heighten of the spatial resolution of the remote sensing image, the precision of vegetation information extract will be enhanced continually.