Telecentric 3D Measurement System And Data Registration, Integration Technique Research

With rapid development of computer technique, The Three-dimensional Imaging and Modeling have been widely studied and applied. Telecentric imaging has the unique property of purely orthographic projection of scene points, and shows the advantage of increased depth of field and low distortion.It also features as constant magnification within the scope of object distance, which could avoid the imaging parallax produced by conventional lenses. Thus, we introduce telecentric lenses for three dimensional shape measurements with fringe projection.The project is one of the content of "three dimensional sharp measurements based on telecentric imaging and general S transformation (E051102)" supported by national science funds and "the research of the key technique of high speed and high precision Three-dimensionaldigital imaging(No.2011125)" sponsored by high level personal funds of Guangdong province.This thesis focuses on three key issues with respect to Three-dimensional shape measurement system based on telecentric imaging, the point cloud registration and point cloud integration, to realize the object reconstruction by meshing express finally.The main content and major contributions of the thesis involve:1. Base on the existing three dimensional shape measurements system with telecentric imaging designed by our team, the research activities on system calibration and 3-D reconstruction technique have been conducted for the acquisition the 3-D data of the object using this telecentric imaging system. In order to demonstration the feasibility of this method, some evaluation experiments have been carried out by measuring an electronic chip with complicated shape.2. In order to placethe range image taken from different view points into a common coordinate system, we propose an alignment algorithm by combining available algorithm ASIFT and ICP to realize the automatic registration process. Comparing with the existing alignment, this algorithm shows the advantage of robust and fast convergence in the process of accurate alignment with low surface overlap. We also conduce some relative experiments to analyze the registration error.3. Under considering the characteristic of the measurement system and the process of registration, we use the concept-"suture line" to deal with the registeredpoint cloud data in order to achieve the data fusion of multiple point cloud data.And then the integrated point cloud data is resampled by using the Moving Least Square method. After that, we can acquire the object reconstruction by Possion surface express. Experiment results indicate that the proposed method performs well in the integration and surface express.