Multi-scale Segmentation Technique In High Resolution Image Information Extraction Application Research

The appearance of high resolution remote sensing ima ge extends the visua l field of thenature,and makes us obtain very abundant information of nature,But the challenge that facesus is how to make use of the data effectively and obtain more useful information throughsome processing．High resolution remote sensing data such as QuikBird and Geo-Eye have alot of characteristics such as spectral,shape,texture and context and so on．Though thetechnology of the remote sensing ima ge classification has made considerable progress,it willresult in not only reducing the extraction accuracy but also making the spatial data redunda ntand wasting the resource,when the traditiona l feature extraction method is applied to the highresolution remote sensing ima ge．Taken Geo-Eye high resolution remote sensing ima ge of Chang′an district Xi′an city asan example,choosing the typica l area and la nd use abundant area as study area,and regardingeCognition ima ge processing software as the platform,the paper carry on the classifica tionexperiment to the study areas according to multi-sca le segmentation．The paper does theresearch by the follow steps:firstly,according to the characteristic of different surfacefeatures types, choosing the optimum scale to segment the area to extract the objects,secondly,constructing the feature extraction system.,extracting the spectral features andshape feature or feature associa tions of the surface feature types．adopting fuzzy classificationto classify to surface feature types,then getting the feature extraction result of study areas．Atthe end, the paper compares and appraises the features extraction result between the method ofmulti-sca le segmentation technology and pixel-oriented(such as the ma ximum likelihoodclassification)．the result indica tes that:The extracted surface features have higher shape andattribute consistency with true surface features when used multi-sca le segmentationtechnology of method,It has higher precision when used multi-sca le segmentation technology of method to classify the high-resolution ima ge,It is so effective to reduce the"Pepper andSalt Phenomenon",The result of multi-sca le segmentation technology of method is moreeasy to understa nd and expla in．...