Research On 3D And Color Texture Reconstruction Technology Of Structured Light Based On Trinocular Speckle

Along with the comprehensive layout of intelligent manufacturing development planning,various virtual and digital needs are constantly accepting challenges.3D shape measurement and reconstruction technology is an important part of the construction of "Metaverse",and can also be widely used in industrial testing,biomedicine and other fields.The 3D reconstruction of structured light has been studied widely because of its advantages such as non-contact,low cost,high precision and fast speed.In this thesis,the 3D and color texture reconstruction technology of speckle structured light at a long working distance are studied,and build a 3D and color texture reconstruction system of three-mesh speckle structured light with high accuracy.The main contents and achievements of this thesis are as follows:Firstly,aiming at the problem that corner detection of traditional checkerboard calibration board is susceptible to noise and illumination,and the ellipse edge blur,the distortion error and eccentricity error of circular array calibration board produced by camera imaging model,a calibration method based on the improved Franklin moment and the iterative compensation of the projection ellipse error was proposed under the circular array calibration pattern.In this algorithm,a three-order gray model is introduced to improve the traditional Franklin moment,and the improved algorithm is applied to the edge detection of circular patterns.In order to reduce the distortion and eccentricity caused by the projection ellipse,the camera parameters are iteratively corrected by distortion correction and back projection matrix.Experimental results show that,compared with the comparison algorithm,the improved algorithm has higher accuracy for sub-pixel positioning of circular pattern,and the proposed calibration method has lower calibration error than the traditional checkerboard calibration method and the circular array calibration method built in Open CV,so the3 D reconstruction accuracy of objects can be effectively improved.Secondly,in view of the problem that traditional feature detection and description algorithms cannot have both robustness and real-time performance and that traditional reliability guided digital image related methods need to manually specify seed points,this thesis combined AGAST algorithm and BEBLID algorithm to propose an improved AGAST-BEBLID feature detection and description algorithm.The proposed algorithm improves the AGAST algorithm’s problems such as the absence of scale invariance and rotation invariance,redundant information in feature extraction and the need to manually specify the decision threshold.The proposed algorithm is applied to digital image related methods for automatic seed point selection and image feature detection and description in color texture reconstruction.The experimental results show that,The robustness and real-time performance of this algorithm is better than that of the comparison algorithm,and it has good effect on 3D and color texture reconstruction of objects,and has good practicability.Finally,a visual speckle structured light 3D and color texture reconstruction software was developed on the Qt platform,integrating various reconstruction modules.The experimental results show that the software can well complete the digital speckle pattern generation,camera calibration,DIC matching,point cloud generation and display and other tasks.