Research On The Classification And Building Extraction Based On WorldView-2 Imagery

Building, mainly including house, bridge, road and large-scale engineering building, is the various forms of place, which people live and carry out production activities. Besides, it is also the subject of artificial nature. In recent years, along with the rapid development of human society, the population explosion growth and urbanization process speeding up, the conflicts between construction land and the farmland have become prominent. It is very meaningful to extract the information of building accurately, quickly and objectively. It can make us know exactly the information about urban development and city planning can be improved and more reasonable. However, the traditional visual interpretation method is laborious and time-consuming and it cannot satisfy the requirements of production and monitoring. How to get the exact information of building semi- automatically or automatically and efficiently is one of the major focus of recent researches.Up to now, World View- 2 satellite is the highest resolution commercial remote sensing satellite launched by Digital Globe company, and it can take multi-spectral images with a resolution of 1.8 m and panchromatic images with a resolution of 0.46 m. It is also the first high-resolution commercial satellite with 8 spectral-bands, except four common bands (blue band-Band2 (450～510 nm), green band-Band3 (510～580 nm), red band-Band5 (630～690 nm) and near-infrared band-Band7 (770～895 nm), and also having four new bands:coast band-Band1 (400～450 nm), yellow band-Band4 (585～625 nm), red edge band-Band6 (705～745 nm) and the near infrared remote band-Band8 (860～1040 nm) for more abundant information. In this study, based on the summary of researches on building extraction with remote sensing imagry and on the basis of in-depth analysis of the spectral characteristics of World View-2 images, unsupervised classification, maximum likelihood classification and decision tree classification were used to classify and buildings extraction the World View-2 images in Yushan, Fuyang city.This paper includes five chapters. First chapter is introduction, mainly on the research background, the civil and foreign research conditions, the structure and organization of this paper and the idea and the technology way to solve the problem. Second chapter is the general introduction of the study area and image data pretreatment. In this chapter, the natural environment and social economy of the study area are generally introduced and also the pretreatment of the image data that includes radiation correction, geometric correction and image fusion. Third chapter is about the analysis and selection of remote sensing image classification feature. Based on the band combination of World View-2 image, the best band combination way was proposed, then after analysis of the image original spectral features and ratio processing of different bands and analysis of principal component, classifications characteristics of different surface features was obtained. Fourth chapter is remote sensing image classification and extraction comparison research. On the basis of in-depth explore of the spectral characteristics of World View-2 imaging, non-supervised classification, maximum likelihood classification and decision tree classification were used to conduct the object-oriented classification of remote sensing image. Last chapter for this study is the summary and prospects, mainly summarized the main research contents and research achievements, also the discussion of the major problem in the research and as well as to provide ideas for the further research.