Research On Enhancement Of Multispectral Remote Sensing Image Based On Fuzzy Principal Components Analysis

The development of remote sensing technique have revealed human being observing the earth and the prelude sounding an aerospace from deep space which found new approach for we know territory, tapping natural resources, studying an environment and analyzing the whole world. Remote sensing digital image processing is a new branch of science that come into being with computer digital image processing techniques and the remote sensing image analyses technology. Meanwhile, remote sensing digital image processing technology which have changed the traditional image processing and recognition mode fundamentally and has created favorable condition for improving technology, high-effect and fast recognition and the processing of multi-source information digital anatomists.The image resolving power increases greatly which due to the measuration surface feature radiation spectrum energy with mark band. But the eye is short of the absolute value concept to brightness, not able to according every wave band distinguishes surface feature. Meanwhile, correlativity that exists is very highly between remote sensing images. Judging from perception, different band is very much similar. As a result, there are a lot of information is unnecessary and repeated consider from the angle extracting useful information. K-L (Karhunen-Loeve) is also called Principal component analysis, is a favorite multivariate statistical method for data compression and information extraction and is also the most useful algorithm in remote sensing image processing.However, it is well-known that PCA, is sensitive to outliers, missing data, it maybe get deformity and error result. At the same time, remote sensing image contained by various surface feature targets, so have fuzziness. Fuzzy theory is a theory that process fuzziness, and other mathematics implement been not equal to it in resolving this kind of problem. According to this problem, this paper then puts forward the concepts of fuzzy expectation, fuzzy deviation, fuzzy variance, fuzzy covariance, fuzzy correlation, accordingly introduces a powerful approach which based on fuzzy statistics to improve PCA: fuzzy PCA.This thesis relies on Chinese Academy of Sciencer (CAS) northeast geography graduate school-ChangChun active fault high remote sensing image processing and interpretation project, and by means of the multi-spectral remote sensing image to analysis that this project provides. It can be explained,that if fuzzy statistics is applied into PCA by making fuzzy sets to participate in decision making, it can restraint noise effectively and enhance main information and raise accuracy and reliability of the decision result. Finally, aimed at characteristic property of the surface feature spectrum, the fake color having been processed. The difference between surface feature is expanded sufficiently, and surface feature recognition is improved greatly.