Research On Segmentation Methods And Application Of City Building Region Extraction From Remote Sensing Image

Remote sensing image has become an important technology for spatialinformation acquiring. The improvement of spatial resolution of remote sensingimage makes image segmentation more important in image analysis. Rapid andaccurate extraction of urban buildings plays an important role in the renewal ofbasic geographic information. The segmentation and extraction of urban buildingarea of remote sensing image are helpful to improve the accuracy and efficiencyof the building range of the buildings. On the basis of building regional feature inremote sensing image, have a research on segmentation correspondingeigenvectors of several typical remote sensing image, focus on texture frequencydomain characterization of building regional, building a texture feature vector.Have an analysis of the basic method and the development direction of imagesegmentation, and summarizes the features of several typical segmentationmethods. Threshold operation speed, suitable for parallel processing, but onlyuseful to image with high contrast between background and target. Edge detectionmethod is suitable for boundary which has gray mutations and has a poor effectof local-gray-mutation image. Region-based segmentation methods make full useof image feature value similarity between elements and spatial correlation, andthey are insensitive to noise interference, applicable to different kinds of imagesegmentation. The integrated segmentation algorithm improves the accuracy ofimage segmentation by applying specific data theory, methods and tools. Giving a summary of the three main types of city buildings area, residentialconstruction area, public building area and production building area. summarizesfour common structure of the three types of building area, determinant, pointgroup type, peripheral and hybrid. Have an analysis of the spectral features,thecontextual features and image texture features of the buildings. Give anintroduction of textual Information Representation. Describe image features ofbuilding area in several typical remote sensing image.According to the image characteristics of the building area, analyzes severalresolution remote sensing image segmentation using the eigenvector and thespectral feature vectors of construction rules and texture feature vectorconstruction method. Using two-dimensional Fourier transform method is thecalculation to the main direction of the building, arrangement of building regionaltexture feature. Through the analysis of the building area of ground object spectralenergy distribution, judge buildings in spatial domain main direction and periodic,the texture feature vector of building construction.Conduct image segmentation experiments on several typical remote sensingimage and construct the corresponding image segmentation feature vector,compares the experimental results with manual interpretation results, analysis ofthe experimental segmentation accuracy, to verify the feasibility and effectivenessof the feature vector.According to the experimental results, analyzes the factors affecting of theimage segmentation process and point out contents that need further study.