Building Height Extraction From GF-7 Satellite Images Based On Roof Contour Constrained Enhanced Stereo Matching

Building height is one of the basic geographic information for urban construction planning and analysis.With the development of remote sensing technology,the remote sensing images taken by satellites provide people with sufficient data for observing the earth and become the basic data for surveying and mapping and earth observation.Estimating the height of buildings from satellite images is still a very challenging problem.With the successful launch of China's first sub-meter precision surveying and mapping satellite GF-7,it provides a high-quality data source for China's surveying and mapping field.However,due to the 26 ° inclination of the forward camera of GF-7 satellite and the angle of view between the forward and backward cameras more than 30 °,it is difficult to match the homonymous points in the area of high-rise buildings,which makes the building height estimation based on GF-7 satellite a challenging work.The traditional method of extracting building height based on DSM is to project the building contour on DSM,extract the elevation of the top and bottom of the building,and subtract them to obtain the building height.However,due to the large difference of viewing angles between GF-7 stereo images,the information on the top of the building is missing in the forward and backward images.There are large holes in the roof area of the building in DSM,which can not correctly estimate the roof elevation of high-rise buildings,resulting in the low availability of DSM based method in GF-7 image.Aiming at the difficulty of matching the roof elevation of GF-7 high-rise building,a contour constraint matching enhancement method is proposed to extract the building height.Generally,the traditional DSM method is used.This paper does not discuss DSM and contour acquisition.For each building,three steps are used to estimate the height of a building:1)First,epipolar image blocks are generated for matching,in which the parallax search range on forward image is determined by using the DSM data and a possible building height range in study area.2)Second,roof contour constrained stereo matching is conducted between backward and forward image blocks,in which the difference of standard deviation maps is used for the similarity measure.To deal with the multi-height problem of podium buildings,the gray difference image is adopted to segment podium buildings,and re-matching is conducted to find out their actual heights.3)Third,the building height is obtained through the elevation difference between the building top and bottom,in which the evaluation of the building bottom is calculated according to the elevation histogram statistics of the building buffer in DSM.Experiments are carried out on GF-7 images of Yingde in Guangdong,Xi'an in Shaanxi and Guangzhou in Guangdong.Compared with the traditional shadow based method and commercial software method,this method significantly improves the accuracy of high-rise building height estimation.In summary,the innovation of this paper is as follows:1)In the aspect of building height extraction,a method based on roof contour constraint matching is proposed to extract building height.This method make full use of the image features of the building in GF-7 image,and focus on the building roof matching.The roof elevation of the building is obtained by matching the roof of the building on the epipolar stereo image pair,so as to eliminate the lack of roof elevation caused by large viewing angle.2)Based on roof contour constraint matching,a multi elevation building height estimation method considering podium building is proposed.In this paper,the mask gray difference is used to segment the elevation surface iteratively.The matching result is evaluated by calculating the gray difference image between the templates and the forward matching results.The gray difference image is updated by operations such as connected area analysis,morphological processing and vector segmentation until the area in the mask is correctly matched.This paper proposes a method to solve the height estimation of multi elevation buildings,and solves the problem of recognition and segmentation of multiple elevation planes in buildings by using mask gray difference.