CIP -CATALOGING-IN-PUBLICATION Dornelles, Thiago de Azevedo Online frame-to-model pipeline to 3D reconstruction with depth cameras using RGB-D information / Thiago de Azevedo Dornelles. -Porto Alegre: PPGC da UFRGS, 2020. 70 f.: il. Thesis (Master) -Universidade Federal do Rio Grande do Sul. Programa de Pós-Graduação em Computação, Porto Alegre, BR-RS, 2020. Advisor: Claudio Jung. 1. 3D reconstruction. 2. Visual odometry. 3. RGB-D Cam-Several challenges in computer vision and robotics involve

more »

... veloping algorithms capable of using partial spatial information to generate a reliable 3D perception of the world. Various breakthrough applicable technologies such as Mixed Reality, Autonomous Robotics, Autonomous Driving, Reverse Engineering, 3D Printing, among others, depend on this research topic to move forward. In order to implement a complete application to build 3D reconstructions for individual objects, this master's thesis presents an online pipeline for incremental 3D reconstruction and 6-DoF camera pose estimation based on colored point clouds captured by consumer RGB-D cameras. The proposed approach combines geometric and photometric matching of data provided by both depth and color sensors through an adaptive weighting scheme that copes with eventual misalignment errors between RGB and depth data. Our experimental results indicate that the 3D reconstructions achieved by the proposed scheme are visually better than competitive approaches.