The 3D Reconstruction Of Single Image Based On Vanishing Point Optimization

3D reconstruction of single image does not face the problem of feature matching in the process of multi-image reconstruction,and it needs less image information.It is a hot spot in the field of 3D reconstruction.It has played a great role in urban planning,UAV and digital media.it has attracted the attention of experts and scholars because of its advantages of less information and convenience,In real life,many objects have cuboid structure,so we choose the object of reconstruction is cuboid.In this paper,the three-dimensional reconstruction is carried out for a single image with a rectangular structure,and the reconstruction is completed according to the camera calibration and the geometric property of the object itself.In the process of camera calibration,we need the vanishing point of high precision.This paper presents vanishing point detection algorithm by use linear feature to get effectively line,we complete improvement of vanishing point precision.First,extracting the line from the image.According to the characteristics of the line of the vanishing point,the line beam of three different directions is grouped,and the linear distribution is satisfied in the parameter spaceaccording to the lines in each direction.The improved weighted regression algorithm is used to eliminate the error line and then use the shortest distance method to get the vanishing point.Then obtain the camera intrinsic parameters according to the nature of the vanishing point.The corresponding 3D coordinates are calculated by calculating the internal and external parameters of the camera and the geometric features of the object itself.Finally,the 3D reconstruction of the object is showed by using the OPENGL.The method effectively eliminates invalid line interference and improve the accuracy of the vanishing point,so as to realize the accurate calibration of camera.Taking the package box and the express box as an example,through the experiment contrast process,the improved vanishing point detection algorithm gets more effective lines and realizes the accurate calibration of the camera.The error of 3D model reconstructed is less than2.5%,and it can achieve 3D reconstruction well.