Semi-automatic Building Extraction From High Resolution Remote Sensing Images

In recent several years, the resolution of satellite image has been improved greatly. High resolution satellite images present the characteristics of the man-made objects more accurate and precise, which results in finer structures and details about the man-made objects. Although the resolution of the satellite image has been improved greatly and even human, vehicles and trees can be recognized unambiguously, the techniques of processing and information extraction from remote sensing image progress slowly. Manual interpret is still the major approach for information extraction, which is time-consuming and labor-intensive and becomes the serious bottle neck to the extensive application of remote sensing image. As one kind of important man made objects, building recognition and extraction is of great significance for the spatial data acquisition, image understanding, cartography and many other applications.Base on the geometric characteristics of the building, this paper makes a deep and comprehensive research on the building detection and extraction from the high resolution remote sensing image. The major contributions are listed as follows。1. This paper executes a deep research in shape representation and description, presents many approaches for region representation and feature description, and demonstrates some important algorithms of great importance in region-based and edge-based object recognition are also implemented and listed.2. In terms of building detection and shape representation on large extent of image, this paper presents two schemas, which are bases on pose clustering and SIFT. Then this paper utilize region adjacency graph to extract building and explores several effective approaches for building boundary fitting and introduce a novel building extraction schema base on graph search. The approach proposed has been implemented in windows platform.3. Finally, extensive experiment on building extraction has been executed, and the advantage and the defect of the approach presented will be discussed. Some potential issue are also been presented, which is in hardly need of been resolved.