Research On Quantitative Analysis Methods Of Hyperspectral Absorption Peaks

Hyperspectral remote sensing,as a technology developed in recent years,plays a more and more important role in remote sensing area.Hyperspectral matching is useful in many hyperspectral applications.In order to realize the hyperspectral classification and other hyperspectral technology,accurate spectral matching algorithms are needed.The hyperspectral curves can be considered as high dimensional vectors,and it has the advantages of both the directional and the numerical.The traditional spectral matching is focused on one of characters.At the same time,because of the noise and atmospheric interference,hyperspectral matching effect is reduced.In order to solve this problem,we focus on the spectral matching method that based on the characteristic parameters of hyperspectral absorption peaks.The structure of this thesis is as follows:(1)We summarize the basic knowledge and research status of hyperspectral matching.Basic knowledge of hyperspectral matching and hyperspectral absorption characteristic parameters are introduced.According to compare the hyperspectral matching algorithms,we know the advantages and disadvantages of them.(2)A hyperspectral matching algorithm is proposed.According to the different absorption peaks,we carry out the spectral matching.First,the continuum removed and normalization methods are used to obtain the spectral absorption peaks characteristics.Based on the new spectral absorption feature parameters and the redefinition of spectral characteristic parameters,the absorption peak characteristic parameters matrix is formed.Finally,we perform the hyperspectral matching in the sight of the similarity of matrix.The performance of the method is verified by two experiments.(3)To address the problem of spectral matching caused by the difference in absorption peak distance when adopting traditional ground objects identification,it brings up an optional algorithm to improve the hyperspectral matching and carries out the spectral matching according to the selected absorption.This algorithm begins with the continuum removal in spectral curve and extraction of spectral characteristic parameters.Then it gradually searches for the cosine distance and euclidian distance of absorption peak according to the maximum likelihood of each standard and to be detect characteristic parameters.Hereafter we conduct the spectral matching according to the selected absorption peak characteristic parameters matrix.The results show that this algorithm can match the best characteristic parameters vector,realizing the selection of absorption peak.After the matching of selected absorption peak characteristic parameters matrix,the error of spectral matching shows a slight decrease.