Abstract


3D reconstruction methods using images captured by cameras are easy to use because they employ only cameras. However, these methods have a problem in that they cannot generate 3D models of regions that have visually few features. In this study, a 3D reconstruction method using a pendulum RGB-D camera has been proposed. This method generates 3D models of regions with few features utilizing pendulum swing motion that oscillates according to the laws of physics. In this method, the environment is first photographed using an RGB-D camera with a pendulum motion, and tracking is performed to estimate the motion parameters of the pendulum such as the fulcrum position and length of the pendulum using only the images that were trackable. The estimated fulcrum position, length, and time when the image was taken are then used to determine the camera poses for all images including images that do not have enough features to be tracked by image processing. Finally, a 3D model of the environment is constructed using all color and depth images.

Powered by JDBS Ver.2.1.0
(C) 2020 Hirotake Ishii