Indoor Scene Reconstruction Using a Rotating Device and Multiple RGB-D Cameras
Zunjie Zhu,Yuxin Wei,Rongfeng Lu,Chenghao Xu,Xiang Le,Bolun Zheng,Chenggang Yan,Feng Xu
DOI: https://doi.org/10.1109/tim.2024.3420266
IF: 5.6
2024-07-19
IEEE Transactions on Instrumentation and Measurement
Abstract:The real-time 3-D reconstruction of indoor scenes is important in computer vision and graphics. Existing solutions mainly use a handheld red, green, blue (RGB) or red, green, blue, depth (RGB-D) camera to perform the task. However, these solutions do not consider information other than visual features to estimate camera poses and require intensive labor to record the scene. In this article, we design and develop a motor-driven rotating acquisition device with three synchronized RGB-D cameras to automate the acquisition and propose a novel method to assist the camera pose estimation by the angle encoders of the motor. First, benefited from the design choice of the device, the camera pose estimation can be restricted in a 1-D space, and motor and visual information are combined together to achieve high accuracy. Furthermore, motor-based loop detection and global optimization are proposed to refine the camera poses and the scene geometry. Extensive evaluations are performed on both virtual and real datasets to demonstrate that the proposed indoor scene reconstruction system outperforms existing solutions regarding both reconstruction efficiency and accuracy.
engineering, electrical & electronic,instruments & instrumentation