The Method Of Object-oriented Building Features Extraction From High Resolution Remote Sensing Images

With the rapid development of remote sensing technology, remote sensing image resolution has been greatly improved and the form, shape, and related information of ground targets can be obtained from remote sensing images. Therefore, the development of high-resolution remote sensing image information extraction technology has very important practical significance. Commonly feature extraction technology based on visual interpretation and pixel gray can not meet the actual needs of the mainly production, while object-oriented object extraction is becoming the direction. Building is one of the most important elements in geographic information and primary elements in cities, towns and rural, which is closely related to human life. Obviously, building information extraction technology has great significance in update of geographic information database and the digital city applications and urban planning and construction.This article first analyzed of the existing research progress including multi-scale segmentation technique, object-oriented information extraction art and high-resolution remote sensing image building extraction technology in the field of high-resolution remote sensing image building information extraction technology, and current achievements and shortcomings were analyzed. Focusing on the problems of high-resolution remote sensing image object-oriented building feature extraction technology, research and analysis has been done in the following aspects:It analyzed the spectral characteristics, geometric features and texture features three aspects of buildings on the image spatial of high-resolution remote sensing images. Introduction was done on parametric equation that may be used in each feature extraction of the process; summary of the building geospatial characteristics was concluded, including characteristics of the spatial location of buildings, adjacent relationship between building and surroundings.Study was taken on multi-scale segmentation system that experiments have done on scale parameter settings in the segmentation process using in different area, different resolution remote sensing images. Multi-scale segmentation parameter settings are related to the scale size, color, shape and the shape of the smoothness and compactness, and the image resolution, the information content of the image layer, the secondary thematic information and the different effects of target objects at the mean while. Experimental results show that the size scale that has relation ship with features size and image resolution; color parameters correspond with surface features internal spectral features; features image geometric features directly affect the shape parameter settings.Summery were worked out on two object-oriented information extraction methods, nearest neighbor classification and decision supported fuzzy classification. These two methods were used for different regions buildings separate extraction taking into account of spectral characteristics of the building, geometrical features, texture features, and spatial topology relationship, then, analyzed the extraction results, evaluated their precision in experiment two separately. The highest accuracy of experimental the results is93.41%, the average accuracy89.7%.This paper summarized the image features of the buildings and achieved a reasonable grasp of the scale involved in the segmentation and used the object-oriented methods extraction of different type of building, which has a certain reference value on object-based processing, analysis and research in remote sensing image.