Characteristics Analysis And Image Classification Of Remote Sensing Images Based On ENVI

The remote sensing images have been applied more and more widely in resources,environment,military and other fields since the 1980s.The classification of the remote sensing image is an important aspect of remote sensing applications.This thesis mainly introduces remote sensing image processing and characteristic analysis based on ENVI,and then studies the classification methods of remote sensing images.The main works includes:(1)The overview of remote sensing.This thesis mainly introduces the remote sensing system and imaging spectroscopy,and then summarizes the development of the imaging spectroscopy at home and abroad.The difference between the multispectral and hyper-spectral remote sensing images is discussed combined with the characteristics between them,then expounds several common storage formats of remote sensing image and discusses the applications of hyper-spectral remote sensing images.(2)Image processing.Remote sensing image processing includes image stretching,image synthesis and color transformation in order to see the image more clearly.The display quality of the image is improved by using gray stretching and histogram equalization methods.At the same time,the feature and boundary of the image can be better distinguished by changing the image color.(3)Characteristic analysis of remote sensing image.This thesis analyzes the characteristics of remote sensing image by using ENVI,including principal component analysis and texture analysis.Principal component analysis can show the main information of the image by using several non-correlated parameters.This method can ignore other band noise and show good results.The texture is a natural property of the image itself.This thesis manly discusses the rules and characteristics of the texture in the image.The texture analysis methods widely used are probability statistical filtering and second order probability statistical filtering.We can analyze the image by contrast,mean,variance and other characteristic parameters.(4)The classification of remote sensing image.Different features are divided into the specified categories by the classification of remote sensing image.Typical classification methods are unsupervised and supervised classification.The difference between the two methods is whether using prior knowledge.The unsupervised classification only uses image samples and determines the classification results by exploring the actual objects after classification.Supervised classification uses a classifier to divide the image into different classes,which is established by characteristic parameters and discrimination function.So the method of supervised classification needs prior knowledge.This thesis mainly uses the Isodata and the K-means to classify the TM image and then classifies the TM image by using the maximum likelihood and minimum distance.The classification results were evaluated by observing and comparing the image classification accuracy reports.