Research On 3D Reconstruction System Based On Structured Light And Stereo Vision

With the improvement of intelligence in production and life,the demand for 3D reconstruction technology in many fields is increasing,and it plays an increasingly important role in machine vision,medical and health,aerospace,industrial production,and automatic driving.In recent decades,the advancement of 3D reconstruction technology has been swift,with a multitude of techniques for reconstructing 3D images being created.Structured light and stereo vision-based 3D reconstruction has become increasingly popular due to its affordability,exactness in measurement,and ability to be adapted to rapid 3D reconstruction.The aim of this thesis is to study how structured light and stereo vision can be used to reconstruct 3D images,and completes the design and construction,calibration,stereo matching and experimental discussion of the 3D reconstruction system.This thesis is primarily concerned with three aspects:(1)In view of the low calibration accuracy of the traditional calibration method and the insufficient effect and performance of the SGM stereo matching algorithm.First of all,this thesis the improved "Zhang’s" calibration method,and the "Nine Palaces" method is proposed for full calibration in all directions,and tangential distortion and error factors are added to the distortion model to further improve the calibration accuracy;secondly,the SGM stereoscopic method is improved.In the matching algorithm,the Census transformation is used to replace the mutual information calculation in the cost calculation step,and the stereo matching algorithm is improved by four-path aggregation in the path aggregation;finally,through experiments,the improved stereo calibration algorithm is better than "Zhang’s" The calibration method reduces the re-projection error by 6.6%,and improves the calibration accuracy.Compared with the SGM algorithm,the improved stereo matching algorithm reduces the error matching rate by 1.1% and the matching efficiency increases by 6.8%,which verifies the improved stereo calibration algorithm and improved effectiveness of stereo matching algorithms.(2)Design and implement a 3D reconstruction system based on binocular structured light,including system requirements analysis,general design and detailed design.On the hardware system,the parameter setting values of the camera and projector in the hardware system are determined through calculation,which provides a basis for the camera setting and the angle and distance setting of each hardware.Secondly,this thesis develops a 3D reconstruction system based on C++ and Open CV graphics library,which has the functions of image acquisition,image processing,stereo matching and 3D reconstruction.The 3D reconstruction system combining software and hardware provides a platform basis for data acquisition and 3D reconstruction.(3)The 3D reconstruction system in this thesis is tested and analyzed,and the completeness of each function of the system is tested according to the requirements analysis and system design indicators.After testing,the 3D reconstruction system in this thesis meets the design requirements and can realize various functions to complete the final 3D reconstruction goal.