Research On Building Extraction Method Of High Resolution Remote Sensing Image

Pioneered and developed in the early 60s, remote sensing technology has made rapid development, enhancing sophisticated images taken by satellites. With the increase of remote sensing data obtained, it is extremely urgent to develop new and more efficient method for remote sensing data processing. Making use of remote sensing image, the man-made features’information of buildings can be extracted. As buildings are the main part of man-made objects in the city, developing an efficient method of building recognition in remote sensing images is of great help for map making and improvement and update of geographical information system, and it is also significant in providing theories and methods for other kinds of man-made objects extracting.This paper, combining color invariance theory and utilizing shadow information in remote sensing image, will extract buildings in the image. It begins with extraction of reed roof of buildings based on definition of specific color invariants and then extract the building shadow, and then it goes on with illuminant direction estimation according to above results, finally it estimates the location of non-red roof on the base of building shadow information and illuminant direction estimation and completes the extraction of buildings by using box fitting method. The experimental results suggest, the approach has accurate detection and the algorithm is fast and reliable, which is feasible.