A Best View Planning Based 3D Full Detail Textured Model Reconstruction Process through DLT and Dense Disparity Mapping Algorithm

• Main Authors: Cheng-Yao Lin, 林正曜
• Other Authors: Chih-Hsiong Shih
• Format: Others
• Language: zh-TW
• Published: 2010
• Online Access: http://ndltd.ncl.edu.tw/handle/75813655841647463533

碩士 === 東海大學 === 資訊工程學系 === 98 === A fast and precise 3D reconstruction process is proposed. It aims to reconstruct a fine detail photo realistic textured digital model of the real world objects. This process provides a method that apply image feature points matching and alignment of consecutive photos. The taking of photos on real world objects has been planned based on a view planning and view sequencing algorithm. So long as corresponding control points of two photographs align with each other repeatedly, we can offer a coarse model of the original real object as a whole. Once the view sequencing process is complete, we can not only acquire key frames for model reconstruction, but also receive enough information for full detail texture mapping of the digital model. The whole experiment procedure uses DLT calibration, disparity mapping algorithm and Delaunay triangulation for model subdivision. The use of DLT stereo intersection algorithm allows not only stereo model extraction, but also facilitates texture mapping of the object details. Experiment results shows that the 3D reconstruction model with full textured mapping . When combined with the best view planning and sequencing theory, its role is more valuable than original thought in 3D model digitization.