Research On Buildings Extraction Technology On High Resolution Remote Sensing Images

The high spatial resolution is one of the most important aspects in high-resolution remotesensing, it has been attached great importance in global military and civil domain for itstremendous military value and economic benefits, at the same time, it is also the backbone ofsystemic development in spatial information industry. Information extraction and targetrecognition is the intermediate links of serving all walks of life by converting data to informationin high spatial-resolution remote sensing it’s also the core technology in the application servicechain of high spatial resolution. The target objects information is various and in greatperformance, also contains abundant spatial structure and distributed information. However, thedevelopment has been restricted for its complicated background and noise information.In this paper, we has analyzed Scale and segmentation problems in image interpretation,designed different processing flow to different objects on the basis of experimental results andfeature information of given objects. The main contents are following:1. The paper introduced existing high spatial-resolution remote sensing satellite systems andapplication fields of high-resolution remote sensing briefly. The paper analyzed the researchstatus and development tendency of interpretation of imagery surface features.2. The paper contrasted four image segmentation methods and analyzed the effect throughexperiments. The paper emphasized the multi-scale segmentation methods and watershedtransform of region-based segmentation, the paper improved region merge method.3. The paper designed a method of building extraction in multi-scale segmentation. Thepaper made deep understanding in several key steps, for concept and election of the best scale,expression of object character and system of classification. At the same time, it has testedaccording to character of test image and intended purpose of research.4. The paper designed and realized experiment in basic of watershed segmentation ongradient image and combination the expression of texture feature. Finally the paper analyzed theresult of the experiment and received favorable effects.