Research On 3D Reconstruction Using Digital Fringe Projection Techniques

The research of computer vision has been developed rapidly in recent years.Three dimensional reconstruction technology is one of the most important applications of computer vision.It has been widely used in many fields such as quality control,reverse engineering,digitalization of cultural relics and so on.Due to its advantages of high speed and high accuracy,3D measurement based on grating projection is a significant branch of non-contact optical measurement technology.In this paper,the digital fringe projection based 3D reconstruction technology is studied theoretically and practically.We analysised two parts of main techniques in panorama measurement including the acquisition of point cloud and the registration of 3d point clouds.And a three-dimensional reconstruction system is developed for experimental verification.The main research contents and contributions of this paper are as follows:The calibration for 3D shape measure model using digital fringe projection is studied.To avoid the phase error caused by the gray gradient of the checkboard,features on the checkboard is usually characterized by the circular centers in fringe projection based measure systems.In allusion to the most common used round-array checkboard,an adaptive feature extraction algorithm based on contour extraction is proposed.A phase self-correcting method based on phase shifting is implemented to project sinusoidal fringes.Phase computation technique based on Multi-Wavelength Phase Shifting is studied.In order to eliminate the jumping error in phase computation,we use two step verification to unwarp phase,obtained more accurate result.A phase screening combined with basic image processing technology is used to partition the effective region,achieved the purpose of denoising in practical measurement.A rigid point cloud registration algorithm is proposed.,Proposed a rigid point cloud registration algorithm.An approach of point cloud classification based on point feature histograms is proposed in this paper.We propose a method of establishing the point feature histograms descriptor to match feature points in different clouds.To reject the outliers,RANSAC algorithm is used.In the fine registration,we apply a dynamic constraint of corresponding distance to improve the traditional ICP algorithm.The experiment shows that our process of registration works fine with higher accuracy and efficiency.A 3D reconstruction system is established based on the above research work.At the end of this paper,we studied how to build a reconstruction system with a camera and a projector,and designed the system supporting software.The software is developed with IDE Visual Studio 2013,provide the basic function modules including image acquisition controller,system calibration,point cloud registration,as well as the utility auxiliary tools including digital image processing tool,point cloud editor,etc.The whole workflow of real-time reconstrution system is implemented.