Study On The Technology Of Building Extraction Based On UAV Aerial Image

In recent years, although the rapid development of UAV mapping,extraction technology of UAV images still is in interactive mode. It cannotextract features fast, intelligently, automatically, which has become one of the bottlenecks restricting the further rapid development of UAV. In this paper, based on high resolution images of low altitudeUAV, the author proposed a morphological theory, which realizes the rapid building extraction and semi automation.This paper describes the acquisition and processing techniques of UAV aerial imagery data. Starting fromextraction techniques of the traditional building UAV, the author generates traditional method of UAV videoon buildingextraction.On this basis, the author began the research on semi-automatic extraction of buildings. First, the author concentrates on the building edge extraction technology, and demonstrates several classical edge detection operators in detail. Through comparative analysis, the author chooses“Canny”as edge extraction operator in this study, what's more, the author improved“Canny” edge detection algorithm. The improved Canny algorithm for building can get better image denoising, and can obtain a more accurate edge extraction.Secondly, through the analysis of the external features of the building, the author classifiedthe buildings beforeextraction, and discusses the importance of building extraction classification beforeextraction. Through the threshold segmentation method,the author can successfully conduct initial segmentation between a building and its surrounding environment, and proposes a binary image noise removal method based on the size selectivity.Finally, by using morphological principle, the author conducted the optimization of repair on the binary images of buildings, and achieved the extraction of building binary image, the edge detection for binary image by edge detection operator. What's more, buildings edge information without any noise pollution can be obtained.Ultimately the semi-automatic extraction of buildings is achieved, good results.