Key Technologies And Software Realization Of 3d Reconstruction Based On Structured Light

Structured light based 3D reconstruction technique which is a technique to obtain information of object surface and reappearing it in computer has been through three decades of development and maturity.The technology is now widely used in the field of measurement and reverse engineering.But the technology has not become popular among small and medium-sized enterprises and individual users,one of the reasons is the high threshold to use the software.Existing commercial reconstruction software is generally expensive,deeply relay on the hardware,and the reconstruction process is not sufficiently transparent to users.Meanwhile,currently the open source reconstruction software is often weak in friendliness and usability.To overcome the factors above and make the 3D reconstruction techniques serve the small and medium enterprises and individual users,based on deep study of 3D reconstruction in the aspects of camera calibration,structured light encoding and decoding,as well as point cloud registration and other key technologies,a fully functional 3D reconstruction software was developed.The main achievements of this thesis are as follows:(1)Asymmetric dot calibration plate was used as the mode object in camera calibration,and the calibration algorithm was implemented with OpenCV.The major processes were automated except that the acquisition of calibration board images requires human intervention.The calibration accuracy was given by test for the method.(2)Gray code theory which was confirmed to be unfastidious in the hardware performance was used to generate structured light coding pattern.According to the coding method of projection area,the scanning and reconstruction process was designed for both direct coding method and epipolar contrast method which simplify the decoding process,the triangulation methods for these two methods were given correspondingly and were applied to the reconstruction and the point cloud registration process.The reconstruction accuracy of the two methods was compared.(3)Labels were stuck to the surface of the object to be reconstructed so that point cloud registration could be realized.A line scan method by detecting the edge of the label was proposed,and was used to determine the center of the mark points,which was compared with the ellipse fitting methods in accuracy and efficiency.The compare result proved that the method proposed was applicable for the detection of the annular labels used in this paper.(4)Proposed Least Error Triangle(LET)method to automatically identify the identity of labels.Based on the quaternion method the point cloud transformation matrix was calculated with the identity matching as the initial condition.(5)Developed a practical three-dimensional reconstruction software based on the use of C++ language on Qt platform.Integrated the algorithm and program into mediator architecture.The actual use confirmed that the software realized the tasks which included camera calibration,structured light pattern projection,camera control,point cloud generation and registration,3D virtualization,etc.