Research And Application On 3D Reconstruction Technique About Building Based On SFM

With the fast development of computer technology and other technologies concerned in recent years,3d reconstruction technique has been widely applied in defense and military fields,medical treatment,urban planning,the cultural relic reconstruction and so on.Because of the increasing demand of obtaining automatically the three-dimensional information of the surrounding environment,the requirement of the practical application of3 D reconstruction technology is becoming higher and higher.Therefore,how to reconstruct the 3D model with high precision rapidly and efficiently has become one of the important research topics in the field of computer vision.To solve the actual application problem,this dissertation researches on the technology of automatic reconstruction of buildings from the view of Structure from Motion based on images.Several key techniques in 3D reconstruction process are researched in this paper,specific studies are listed as follows:Firstly,in view of the characteristics of the acquired image,a method based on scale invariant feature transform(SIFT)is used to extract the feature points of the images and match them,on the basis of the research of image feature point detection and matching technology and complete the the establishment of the corresponding relationship between image pairs,correspond the imaging points in two different images of the same physical space point one by one.Then,a dense matching algorithm based on region growth is introduced to achieve the matching points of large orders,after obtaining the initial matching points,.Secondly,an improved random sampling consensus(RANSAC)algorithm is proposed.A sample pre inspection is introduced into the model pre inspection,incorrect sample sets and unstable temporary models are discard quickly in the improved RANSAC algorithm.The number of iterations and the threshold of distinguishing internal and external points are adjusted adaptively to improve the accuracy and robustness of the algorithm in the process of calculation.The number of consistent set points and the epipolar distance average as the criterion of optimal consensus set at the same time to get the result of the smallest error under the condition of obtaining as many interior points as possible.Also,the improved RANSAC algorithm is verified and compared with the traditional RANSAC algorithm in this paper.Again,matching point purification and estimation of fundamental matrix is achieved byimproved RANSAC algorithm.The improved RANSAC algorithm is combined with 8 point method,the initial matching point set obtained by SIFT algorithm is refined by iterative computation to obtain a set of matching points and fundamental matrix with high precision.Experimental results show that the method proposed improves the speed and precision of the method.Then,the technique based on Kruppa equation is adopted to solve the camera internal parameters.The quadratic nonlinear relationship between the two images,which is absolute conic with invariance and the concept of epipolar transformation,is used to solve the Kruppa equation,camera internal parameters is obtained by solving the Kruppa equations directly.Finally,three dimensional reconstruction of building is implemented.The experimental results of previous research work is utilized to restore the relative pose of the camera and 3D scene information,and the obtained 3D point cloud is triangulated to obtain a complete 3D model to realize 3D reconstruction.