Space Frequency Domain Analysis Of High-resolution Image And The Extraction Of Building

With the rapid development of science and technology,society has gradually become informatized and automated.The development of disciplines is even more so rapid.The sources of remote sensing data are increasing and the spatial resolution is getting higher and higher.Therefore,it is necessary to use these increasingly the popular images.Data,the information needed for social development,has become a hot topic in photogrammetry,remote sensing,and even computer vision.As one of the important signs(buildings)of urban development,human living standards,etc.,its information extraction is the focus of current research hotspots.However,features such as features,textures,and edge information in the current high-resolution remote sensing images are extremely rich compared to traditional remote sensing images,which also makes it possible to perform traditional image building extraction analysis by analyzing spectral features of the pixels.The demand is difficult to satisfy.This is one of the difficulties in the large-scale automated reference of high-resolution remote sensing images.The human visual attention mechanism can directly find the existence of buildings from very complex maps.Therefore,we from the perspective of the visual attention mechanism(significance),and this paper uses the characteristics of the building itself as the low-level features and uses the multi-feature information template as a priori.Knowledge guidance,the extraction of significant buildings.The saliency method in this paper is different from the conventional saliency method,which is mainly reflected in the fact that the prior knowledge background template not only adopts color information but fully considers the characteristics of the building itself,and adopts features such as color,position,and texture as feature information;In order to fully achieve the ideal effect of the template,this paper makes use of the shadow information of the building to determine the scope of the template and then determine the scope of the template.This makes the algorithm more accurate when calculating the saliency;it is achieved through two methods of principal component analysis and complete information analysis.It can both suppress redundant information and highlight the complementary effects of building dominance.However,since the prior knowledge cannot be developed like the human brain,the saliency method of this paper can only extract relatively discrete buildings.In order to compensate for the disadvantages of this method and avoid the use of traditional pixel-based spectroscopy,this paper proposes and uses efficient spectral feature information to extract building areas.The main use of the frequency spectrum analysis of buildings is to find the main center frequency of the building and the center frequency of the high frequency edge in the frequency domain.At the same time,the main direction of the building area is obtained by using the characteristics of the vertical spectrum of the polygonal spectrum and the peak pattern of the section line of the building area.In the main direction,the center frequency is filtered and inverse transformed to obtain a spatial domain building extraction map.