Study For Automatic 3D Reconstruction Based On The Double Focal-length Structured Light System

The traditional methods of 3D reconstruction by structured light system uses single focal-length of camera,the defects are the measurement system can't obtain the details of the object when the focal-length of camera is short,and the measurement process is inefficient when the focal-length of camera is long,because the camera angle is smaller.To solve these problems,the structured light system with double focal-length is proposed,the short focal-length measurement system can fast measure the object,and the long focal-length measurement system can additionally measure the details of the object,to improve the expression ability of object in the detail area.The main researches of this study are as follows:Firstly,the structured light system with double focal-length has been built,and several key technologies of the system are studied,such as Phase Calculation,Phase Unwrapping,System Calibration and 3D Reconstruction.Secondly,the study proposes the two ball calibration method to calibrate the relationship of the platform and projector coordinate system,and taking advantage of the result to stitch the multi-view point cloud by transforming the coordinates of the point clouds from the projector to the platform coordinate system.Thirdly,there are a large number of noise and background points after the arbitrary viewpoint being measured.So the study proposes the region growing segmentation algorithm to recognize the target object and remove the background and noise points.Fourthly,the long focal-length measurement system measures the details of the object.It should be merged with the point clouds by the short focal-length measurement system,so the improved ICP algorithm is proposed to realize the precise registration of point clouds,and the redundant point clouds are removed.The method has realized the accurate and non-redundant split of the long and short focal point clouds,finally gets the different resolution with different surface of the object.Finally,the feasibility and effectiveness of the proposed automatic 3D reconstruction of unknown objects method is verified by the reconstruction of several objects.