Building Extraction And Occlusion Information Restoration In Oblique Remote Sensing Images

With the continuous expansion of the demand for the stereo information of the ground features in the field of remote sensing image analysis,not only the top surface information of the ground features must be obtained from the top view,but also the other elevation information of the ground features must be obtained from other angles,based on the tilt photography technology.Remote sensing images provide the possibility of three-dimensional information analysis of ground objects.The oblique remote sensing image breaks through the limitation that the traditional front-view remote sensing image can only obtain the top-view angle information of the ground object.It can simultaneously collect remote sensing images from different angles through multiple sensors mounted on the pan/tilt.Making full use of the three-dimensional information of the building features in the oblique remote sensing image can not only extract and analyze the missing facade information in the blind area of traditional front view image shooting,but also provide information support for subsequent applications such as the accuracy of the 3D reconstruction technology of buildings.That is to ensure the completeness of its three-dimensional information extraction.In turn,it can improve the application capabilities of building three-dimensional information in urban construction planning,traffic control and other civil aspects,and it can also provide information guarantee for various emergency situations such as disaster monitoring and assessment.It can be seen that the technology of information extraction and analysis of buildings based on tilted remote sensing images has excellent application prospects.Because the building target is often occluded by other ground objects in the tilted remote sensing image,and the occlusion will cause the loss of part of the building information and affect the integrity of its stereo information,the occlusion situation is considered to be the information of the building in the tilted remote sensing image.An indispensable prerequisite for extraction and analysis.First of all,for the problem of building target extraction and occlusion area determination,this paper proposes a predictive instance segmentation method for tilted remote sensing images,which is oriented to various occlusion scenes in tilted remote sensing images for predictive instance labeling of sample data,and then Based on the extraction of the building target and the occluded area in the image,the outer contour prediction of the occluded part of the building is realized,which improves the accuracy of the positioning of the building and the occluded area and the completeness of information extraction.Furthermore,based on the extraction results of individual targets and occluded areas of buildings,this paper proposes an occlusion repair model based on edge constraints to solve the problem of missing information in the occluded parts of the building.The selection strategy of the mask type and model constraints are analyzed experimentally,and the image restoration model based on threshold convolution is improved.The occlusion restoration model based on edge constraint uses the edge information of the outer contour of the occluded part of the building to effectively complete the targeted information restoration of the missing area of the target building in the tilted remote sensing image,and then obtain the complete structure information of the target building And texture features provide reliable information support for a series of subsequent applications of building three-dimensional information.