Study On 3D Reconstruction Of Object Shapes Based On Monocular Vision

Three-dimensional reconstruction means to create digital models of objects or scenes in space which are suitable for computer to process.It's a key technology in expressing the objective world using digital models.With the rapid development of 3D printing and virtual reality,how to efficiently realize a three-dimensional reconstruction of objects is not only of high scientific value,but also will bring huge economic benefits.In this paper,we propose a low-cost three-dimensional reconstruction system based on monocular vision,which requires very little hardwares besides a web camera,a hand-held red linear laser and three spatial background boards that are pairwise orthogonal.The operator holds the linear laser and scans the object placed on the background boards uniformly and the camera shoots the process of scans.By handling images,we can extract the position of laser lines.Based on spatial constraints between the laser plane and the projection rays,we can calculate three-dimensional coordinates of the object's surface.A three-dimensional model of object can be built by a series of operations such as filtering,reconstruction,visualization and so on.The main contributions of the paper are as follows:1.Propose a three-dimensional reconstruction method based on monocular vision and spatial orthogonal constraints,which can reconstruct the non-texture objects accurately.2.Use the model of Markov random field to fuse the multi-layer point cloud data.This method can improve the unsmoothness of reconstructed surfaces caused by uneven scanning speed.3.Use the PSO algorithm optimizing the camera parameters.It can reduce errors caused by no calibration of the camera and make the reconstruction results more precise.