Research On Building Height Extraction Method Of Satellite Video Images

With the advent of video satellites,satellite video images have entered people's daily life as a new type of remote sensing data.At present,the research on satellite video images is mainly in the fields of super-resolution reconstruction,dynamic target tracking and detection,and there are few researches in other application fields.The building height extraction is the core technology of the city's 3D reconstruction and other related applications.The traditional method uses the remote sensing stereo images to complete the building height extraction.Compared with stereo images,the height information in the video image is more abundant and therefore more suitable for extracting building height.In this paper,combining with the characteristics of video satellite imaging,we have carried out a study on the building height extraction method of satellite video images,which mainly includes the following three aspects:Firstly,the applicability of satellite video images in stereo matching algorithm is studied and analyzed.According to the video satellite imaging mechanism,the problem of image area misalignment in the original video images is analyzed in both theoretical and experimental perspectives.To solve these problems,video image stabilization technology and rotation correction strategy are used to preprocess the image,and a corrected image satisfying the basic requirements of stereo matching is gotten.At the same time,the corrected image is further analyzed and the priori parallax constraint of the image is given,which provides strong support for the stereo matching algorithm.Secondly,the regional stereo matching algorithm was studied.In order to solve the problems such as more mis-matching and unclear edge of disparity map in traditional stereo matching algorithm,this paper focuses on matching search space constraint and window matching strategy,and proposes a multi-window regional stereo matching algorithm based on double parallax constraint,which improves the stereo matching accuracy of video images.Finally,the research on the method of extracting image building height was carried out.In this paper,block adjustment is used to optimize the Rational Function Model of satellite video images,and the position accuracy is improved.Then,the construction of the building height is estimated by using the photogrammetry space forward intersection technology.Experiments show that the building height extracted by this method has high accuracy and can be used for the 3D reconstruction of urban area scenes.