Applied Research Of Extracting Buildings Imformation Based On Object-oriented Method

With the continuous development of space technology, the bands of the remote sensors which are used in the work in space remote sensing activities have been fully expanded, the spatial resolution of remote sensing images is improved rapidly, while the amount of remote sensing image data is also increased into a geometric ratio. High spatial resolution remote sensing images have been widely used in surveying, agriculture, forestry, urban planning, land and resources management, geological and mining exploration, military and other fields. However, correspondingly, the remote sensing image processing and information extraction technology develop slowly in relatively. Image classification based on image spectral information has been developed for many years which has made some progress, but the effect of the interpretation is still not good enough. This restricted the wide application of high resolution remote sensing images severely.Object-oriented remote sensing information extraction technology is developed only in recent years as a new method for remote sensing image interpretation. It is different from the past image interpretation based on image pixels, which is based on the object of pixels in the image, then extract information by the Characteristics of each object. In this paper, I will extract building informationion from QuickBird image, the study point is the application of the object-oriented method which is used in urban building imformation extraction in the target, and then I compared it with results those come from the traditiona imformation extraction based on image pixel. Finally the most appropriate method for building information extraction is selected. The main works of this article are as follows:(1) Introduced the process of remote sensing image preprocessing brieily. To the experimental data of this paper, i described the theoretical basis of remote sensing image fusion and several commonly used algorithms in detail. Using the method of Gram-Schmidt, I fusioned the multi-spectral image and panchromatic spectral images of QuickBird images.(2) This paper introduces the principle of image scale segmentation and several commonly used methods of image segmentation, and study the segmentation results which come from segmentation parameters in the process of remote sensing image multi-scale segmentation. This paper analysis the basic features of building target in the high spatial resolution remote sensing images and set up knowledge base of the building target features, and two methods to extract building imformation from the segmented images.(3) In ecoginition platform, extract buildings imformation from the QuickBird images based on Object-oriented method, then compare the result with another result which comes from the method of unsupervised classification. The results prove that the object-oriented method of extracting building information has more advantages.