The Comparison Among Fusion Algorithms Of Landsat8 OLI Remote-sensing Imagery And The Analysis On Adaptability Of The Classification Of Land Use

With the progressively increasing of booming remote sensing,its demand for Earth observation technology and polyphyletic remote sensing image data with high resolution,hyperspectral, multidate and multi-platform has gotten a fast incresing,and those image data have been extensively applied in various fields of geonomy.The ways of acquiring polyphyletic remote-sensing image are widely divergent,and remote sensing data from different sources have wide difference in spectral resolution, radiation resolution and space resolution,which lead to different adaptability of algorithms when using different remote sensing data to deal with the specific image.Presently, among the earth observation satellites for civil-use, Landsat satellite is the one with the widest application fields and strong applicability all over the world,and its images have been an important source of information for surveying, evaluating and supervising of environment and resources,and the sensors named OLI and TIRS,which are carried by Landsat satellite,have gotten a readjustment in band design and spectral resolution compared to previous TM and ETM.However, current studies aimed at the adaptive analysis about Landsat data, especially the image fusion and image classification which has gotten a relatively wide application for Landsat8 image,haven’t gotten much more attention.Aimming at Landsat8 OLI data,the author chose Mei county of Shanxi province whose elements of land type are relatively comprehensive as the sample region.Firstly, taking some preprocessings on image data,including data reading, image intensification, radiometric calibration, resampling and image cropping,and so on,then acquiring the best band combination B245 by the way of OIF(Optimum Index Factor) factor extraction algorithm under the basis of the analysis on the statistical characteristic of the image band in research area.Secondly,the author used some methods including PCA transform fusion algorithm,HIS transform fusion algorithm, Algorithm of color standard change(Brovey),GS transform fusion algorithm, Wavelet fusion method(Coiflet、Daubechies、Haar、Symlet) and Wavelet Transform combined with traditional method(Wavelet-PCA、Wavelet-HIS) to conduct the fusion experiment on B245.Then the author did a qualitative average from 2 aspects,the spectral resolution and space resolution of fusion images;Then selected some indexes covering standard deviation, comentropy, definition, spectral distortion, deviation factor and correlation coefficient and did a quantitative assessment from 3 aspects, the information of image brightness,space and spectrum;In the end,the author found that among those 10 fusion method, PCA transform fusion algorithm was the best one for Landsat8 image in this area.Finally,the author used object-oriented classification method and the other 3 methods including method of maximum likelihood, support vector machine and BP neural network method to conduct classification experiments.In order to do a quantitative assessment on the classification result,the author refered to some data for this paper including SPOT images of Mei county and land use data from Chinese Academy of Sciences whose plotting scale is 1:100000,and did a visual interpretation under the basis of sample survey,by doing so, the author used the result to classify the land-use type of Mei county into 7 types,as follows:cultivated land, garden plot, forest land, settlement place and land for mining and industry, land for transportation, water area, bare land.By giving a qualitative and quantitative analysis on classification accuracy of every method,the result showed:all the classification accuracy showed a higher degree on the whole,and their ranking from high to low was as follows:support vector machine> method of maximum likelihood> object-oriented classification method> neural network method,those 4 classification algorithms had a big difference on classification accuracy inside the different land-use type.In the end,the author summarized the existing disadvantages of this paper based on the reference about the domestic and foreign research tendency and gave expectation for the further exploration on this research.