The Stitching Technique of Three-Dimensional Point Cloud by Using Feature Points

• Main Authors: Yu-Wei Lin, 林育葳
• Other Authors: 吳明川
• Format: Others
• Online Access: http://ndltd.ncl.edu.tw/handle/5gjqh2

碩士 === 國立臺北科技大學 === 自動化科技研究所 === 105 === The stitching technology in the 3D image reconstruction is very important. The stitching accuracy between the point clouds of different views is improved, and then the accuracy of 3D image reconstruction can also be guaranteed. In addition, scanning the object by dual laser can alleviate the occlusion problems for grooved or steep profiles in 3D shape scanning by single laser. The captured point clouds are further extracted by using the Harris 3D key points to reduce the computation efforts. Furthermore, the Fast Point Feature Histograms (FPFH) is used to get corresponding points and the pose estimation is successively applied to find the point clouds attitude transformation for different views. Finally, the experimental results show the proposed methods can effectively increase the stitching accuracy and reduce the computations time comparing to the traditional Iterative Closet Point (ICP).