Fusion Of Spaceborne VHR Optical And SAR Images For Building Height Quick Extraction In Big Urban Areas

City is an important area of remote sensing monitoring.Buildings,the essential part of urban area,whose distribution and development are the research focus recently.The height of buildings,an important index of the urban longitudinal development,which is the similar important content of urban remote sensing monitoring.Currently,high resolution optical image and synthetic aperture radar image are extensively applied to extract the building height.This two kinds of RS system whose image itself and the method of extraction all have their own advantage in the process of application.However,because of some special factor,there are always some defects.For instance,when use the optical image for the height extraction,which mainly based on the shadow extraction and the imaging geometric relation.But the satellite parameter of some high resolution image cannot be acquired and make the use of method difficult.For the SAR image application,whose main problem are the imperfection of image feature and the low extraction efficiency.In view of the existing defects of the two kinds of image,in this paper,we investigated the joint use of the spaceborne high-resolution optical images and SAR images to extract building height information quickly in big urban areas.This approach,which make up the flaws for each other,give consideration to the accuracy and efficiency.The main research contents are as follows:(1)In the aspect of building shadow from the high resolution optical image,in this paper,we used a method which combined the supervised classification with unsupervised classification.The joint use of support vector machine and morphological shadow index can eliminate the need for amounts of process which to determinate the feature threshold.The method not only improve the efficiency,but also meet the accuracy requirements.The shadow length is automatically calculated and recorded after the shadow is acquired.(2)In the aspect of building height extraction,we extract the building height from high resolution SAR image by using of model matching technique.Before that,we research the imaging mechanism and image feature of buildings.First of all,we establish the geometry model by analyzing the building image feature.Then,the matching function between the geometry model and the SAR image feature is defined.Finally,the genetic simulated annealing algorithm is applied to iterate and find the best result.(3)In the aspect of the combined use,we choose some buildings whose image feature are intact,to extract their height and shadow length.Then,we can obtain the other heights based on the simple linear regression analysis.The study area of this paper is a residential area in Toutun River District of Urumqi City.The method of this paper achieved through create computer programs.We choose five buildings as samples in study area and extract their shadow length and height,then we establish linear relation.We choose twenty-four buildings uniform in study area as experimental subjects,then use the method of this paper extract the height and verify the result through total station.The results which absolute error and relative error lower than 4m and 17% respectively are satisfactory.