Research And Implementation Of 3D Reconstruction System Based On Multiple Images

With the widely use of digital multimedia and the improvement of computer processing speed and storage capacity,3D reconstruction applications also promoted in the practical field.Including like digital medical treatment,virtual reality,the cultural relics,historical building protection and film and television production(3D),education and so on many aspects,which involved the core technology is based on the 3D reconstruction of the image.However,because in the practical application of always could not reach the requirements of speed and precision of the ideal,therefore,3D reconstruction based on image technology has been the focus of research scholars.This article is under the condition of the reconstruction speed and poor precision of 3D reconstruction technology based on image to do some research.First in this article,through reading a large number of both at home and abroad in 3D reconstruction based on image technology knowledge,sort out in this paper,we study the technical route and key steps,including camera calibration,3D model,3D model of optimization and texture mapping the four main key steps.Then,the paper introduces the world coordinate system,the camera coordinate system and image coordinate system and the transformation of the relationship between the three coordinate system,then explains the traditional camera calibration algorithm,and then,based on the principle of camera imaging and the world coordinate system,the camera coordinate system and image coordinate system transformation relations,use the DLT algorithm,Tasi algorithm,ilf plane template method for experiments,finding the internal and external parameters of the camera and analysing the results and getting conclusion.According to the nature of the three kinds of algorithms and the actual situation of this paper,chosing ilf plane template method obtained result for projection matrix.Then,using the Grabcut algorithm to extract the foreground object image,the image binarization processing,systematic scan images until you hit the first non-zero point,to it as a starting point to track his silhouette.marking the boundary pixels,when the outline of the complete closure,return to the previous position until find new components,after finding out all the connected region,according to the contour of the object size,kicked off the other connected area,retained the contour of the object.To get rectangular line scanning contour image,according to the principle of imaging and 3D coordinate system transformation to get the geometric objects point coordinates,has been further AABB bounding box of the object,then the surface of MC algorithm is adopted to subdivise AABB bounding box of the object for obtainingthree-dimensional model.Then,the paper introduces the triangular mesh optimization related vertex deletion method,the improved triangular folding method,edge collapse algorithms,in the light of the actual situation of this article chose the edge collapse algorithms to optimize 3D model,so that the accuracy is not affected,improve the speed of drawing.Finally,in order to get realistic three-dimensional 3D images,this paper introduces the texture of the relevant knowledge,and the principle of texture mapping and commonly used parametric equation of cylinder is explained in detail in this paper,by using the parametric method and the detailed steps of OpenGL rendering drawing,visibility rest use another algorithm can solve the problem of the barrier in the process of texture mapping,using average method to solve the one vertex texture point values corresponding to multiple problems,finally completed the texture of 3D model,the reconstruction of the final results are obtained.In order to further validate the method the effect of reconstruction,the huawei mobile phone boxes of the laboratory is experimented for reconstruction,experiment proved that this reconstruction method has achieved good effect.