Research On Three-dimensional Reconstruction Technology Of Structured Light Based On Binocular Stereo Vision

In recent years,with the rapid development of machine vision technology,how to reconstruct the 3D information of the object from the 2D image has become an important research topic in the field of machine vision.Traditional three-dimensional measurement means are mainly contact point-by-point measurement,due to its low efficiency,high cost and other disadvantages have gradually failed to meet the needs of people real-time measurement.In comparison,optical three-dimensional measurement technology has been widely concerned with the advantages of fast speed,high precision,large measurement range and simple structure.Now,the technology has been widely used in quality inspection,mechanical manufacturing,reverse engineering and other fields.In the field of optical three-dimensional measurement,binocular stereo vision and structured light are widely used as two commonly used schemes,but both of them have some limitations and shortcomings.Binocular stereo vision technology is affected by texture information and lighting conditions.If there is less texture information in the scene or the lighting conditions are not ideal,then the reconstruction accuracy will be adversely affected.In structured light technology,projector calibration accuracy is low and prone to nonlinear errors.This thesis adds structured light projection coding pattern on the basis of binocular stereo vision,and uses phase features for stereo matching to build a structured light 3D reconstruction system based on binocular stereo vision.The research contents of this thesisr mainly contains the following parts:First,the method of phase acquisition is studied.The calculation of phase consists of two parts: wrapped phase acquisition and phase unwrapping.In the aspect of getting the wrapped phase,this thesis mainly introduces the principle of phase shift method,and on this basis,proposes to use Hilbert transform to calculate the wrapped phase,which improves the projection efficiency.In the aspect of phase unwrapping,the principle of phase unwrapping and the advantages and disadvantages of multi-frequency heterodyne method and Gray code method are deeply discussed,and Gray code method is selected as the phase unwrapping method.Aiming at the problem of edge hopping error in the phase unwrapping process of Gray code,the tripartite phase unwrapping method is adopted to solve the problem.Second,camera calibration and stereo correction methods are studied.Camera parameters are one of the important factors that affect the reconstruction accuracy.Based on the keyhole imaging model,this thesis introduces the transformation relationship between each coordinate system and the causes of camera distortion.At last,Zhang Zhengyou checkerboard calibration algorithm was used to calibrate the camera.For binocular systems,there is an Angle between the imaging planes of the camera.In this thesis,the principle of stereo correction is studied,and finally the Bouguet algorithm is used to correct the camera imaging plane.Third,the stereo matching algorithm is studied.In this thesis,several commonly used stereo matching algorithms are introduced,and the SSD-based phase matching algorithm is used for the data distribution characteristics of absolute phase.The method of combining coarse matching and fine matching is used to find the matching points,and then the matching points are fitted three times to make the matching accuracy reach sub-pixel accuracy.The experimental results show that the algorithm still has high reconstruction quality in the face of complex environments.Fourth,the construction of binocular structured light 3D reconstruction system.According to the actual needs,a 3D reconstruction system is built,and the software algorithm part is developed by using Visual Studio 2017 and MATLAB.Finally,through the analysis of the reconstructed point cloud data,the reconstruction accuracy of the system is 0.016 mm.