Research On Remote Sensing Building Detection Technology Based On Visual Attention Computing Model

The analysis and interpretation of remote sensing building image is of great significance in land planning and management,dynamic monitoring of natural disasters,monitoring of land environmental pollution,and military dynamic monitoring.Aiming at the complex scene of large field of view in suburban area of remote sensing building image,the sparse distribution of building targets,the distributed target of suburban buildings,and the high-efficiency detection,this paper proposes a remote sensing suburban building area detection technology based on visual attention computing model.The key contents of the research are as follows:Firstly,a preliminary screening algorithm for complex remote sensing buildings based on integral graph model is proposed.Aiming at the complex texture structure of building area and the requirement of high-time detection,this paper constructs a template of building area based on texture integral map to construct gradient integral map,so as to obtain the preliminary segmentation binary map of building area and realize the rough screening of building area.Firstly,the gradient integral map of the boundary of the building area is calculated,then the template of the building area based on the gradient integral map is obtained,and finally the preliminary result map of the building area screening based on the texture integral map is obtained.Secondly,an automatic detection algorithm based on context-aware visual attention computing model for building candidate extracts in remote sensing images of suburbs is proposed.In view of the characteristics of complex and confusing scenes,sparse distribution of targets and distributed objects in suburban buildings in large field of view remote sensing building images,this paper achieves efficient screening of suspected areas of buildings by combining the model of fast visual attention extraction and context perception.In this paper,an algorithm for the context of candidate areas and visual attention fusion is selected to achieve efficient screening of suspected areas of buildings.Firstly,the primary candidate regions are obtained by using the single-scale local characteristics of the saliency of the primary features at the bottom.Then,the visual saliency candidate regions are obtained by using the principle of structural similarity and global multi-scale focusing.Then,the context Bayesian network is used to extract the distributed targets in the building areas.Finally,the global and local saliency are integrated and the global saliency is integrated.And local saliency to get the final target building screened saliency map.Thirdly,this paper designs an accurate identification algorithm for building candidate areas based on the characteristics of multi-class complex high-rise structures.In view of the characteristics of the multi-variable and sparse distributed scenes in building area detection and the complex scenes in suburban buildings,this paper proposes a method of false alarm elimination based on multi-class and multi-class high-rise structure features and single classifier identification model,so as to identify the final building area.This model makes up for the weak robustness of single feature combined with classifier model in complex scenes.It not only completes the reliable detection of building areas,but also realizes the accurate detection of buildings with high difficulty and sparse distribution.Finally,based on the above three key steps,this paper constructs a complete suburban building detection system based on visual attention computing model.Write a few sentences in which the experiment shows that.It realizes fast and accurate detection of suburban buildings in large field of view remote sensing images.In this paper,with the support of a large number of remote sensing image data,the efficiency and validity of the key technologies and the overall algorithm framework are verified and analyzed by using multi-angle scale,multi-level and multi-level experimental means.This paper uses multi-level and multi-angle experimental verification means to validate and analyze the proposed model and method,realizes the rapid and accurate detection of suburban buildings in large field of view remote sensing images,and achieves the high-efficiency and accurate detection effect of suburban remote sensing buildings under complex large field of view conditions.To provide theoretical methods and key technical support.