Research On Real 3D Model Reconstruction

Three dimensional reconstruction technology is widely used in digital city and archaeological field,the purpose is to restore the real scene and better protection of cultural heritage.At present,stereo vision technology and laser scanning technology are commonly used in 3D reconstruction.The stereo measurement system can obtain high resolution image information,but lack of spatial distribution of the point cloud.The laser scanning system can obtain dense high precision point cloud information,but the texture information is far less than the image resolution.If it could be a combination of both,learn from each other,we can achieve the perfect unity of model accuracy and the appearance of texture,to achieve true three-dimensional model reconstruction.This paper combines the advantages of two methods,through registration and fusion to achieve high accuracy 3D reconstruction with real texture information.One of the key points of this paper is how to realize 2D-3D data fusion automatically.In order to solve this problem,some mark points are set up in experimental scene.This is a necessary condition for laser scanning point cloud splicing.If the stereo measurement system can also use these signs,we can achieve the unity of the two coordinates.According to the spatial relationship of the mark points in two systems,the 2D-3D data fusion problem is transformed into a 3D-3D data fusion problem.Calculating the translation matrix and rotation matrix of the two systems and performing coordinate transformation to get corresponding information between point cloud and stereo image.In order to achieve automation,an automatic extraction algorithm is presented here based on nuclear line and space constraints to calculate the coordinates of mark points.For large scene,multi-view data acquisition method is adopted to achieve real 3D model reconstruction.For contrastive study,two methods are used to reconstruct the model.One method is to reconstruct each site cloud separately.After the Delaunay triangulation and texture mapping according to the corresponding information,the single station with texture model is obtained.Then,the adjacent station model is concatenated according to the spatial relationship of the marker balls.Finally a complete true 3D model is obtained.The second method,first get colorful point cloud with corresponding information,then match point cloud according to the spatial relationship of marker balls and use ICP algorithm to get better registration.Finally obtain a complete 3D model after Delaunay triangulation.Two methods both need to point cloud triangulation,while traditional triangulation algorithm takes a long time and the quality is not high.This paper uses a Delaunay triangulation algorithm based on Voronoi optimization,greatly shorten the program execution time.In this paper,the object of the experiment is the Century Square stone carving and Einstein status of Wuhan University.The experimental results show that the two methods can be used to obtain the real 3D model with high accuracy and real texture information,while the texture information of method one is more refined.The experimental results verify the feasibility of automatic extraction algorithm of center coordinates and optimized Delaunay triangulation algorithm,and it can achieve the fusion of stereo image and laser point cloud data,finally complete the reconstruction of true three-dimensional model.