2D Features as Initial Guess of ICP Algorithm for 3D Registration

• Main Authors: Lin, Chia-Chen, 林佳蓁
• Other Authors: Lin, Chien-Chou
• Format: Others
• Language: zh-TW
• Published: 2015
• Online Access: http://ndltd.ncl.edu.tw/handle/38644578996633436026

碩士 === 國立雲林科技大學 === 資訊工程系 === 103 === This thesis presents a novel 3D registration method which takes features on Bearing Angle Images as initial guess of ICP (Iterative Closest Point) to enhance the quality of registration. The proposed method consists of five steps：(1) transforming a 3D scan into 2D Bearing Angle Images, (2) extracting features from the 2D images by SURF (Speeded-up robust features), (3) finding the corresponding 3D point pairs with respective to the 2D corresponding pixel pairs by the reversed mapping function of bearing image, (4) calculating translation matrices of the corresponding points and (5) finding the best transformation between two point clouds by voting and adopting the best transformation as the initial guess of ICP. In this thesis, there are 6 different kinds of models with different sizes applied on registration in the experiments. On efficiency, due to the less time spent on finding correspondences, the initial guess of ICP not only greatly decreases the time cost of ICP but cut down the iteration times of it to 86%. Furthermore, taking features on Bearing Angle Images as initial guess of ICP also increases the robustness on larger angle diversity up to 45 degrees.