Accurate, Real-time, and Dense Surface Reconstruction of Dynamic Objects using Multiple RGB-D Cameras

Kyoung-Rok Lee, Haleh Azartash and Truong Q. Nguyen

Video Processing Lab

University of California, San Diego

 

Abstract

In this paper, we propose a real-time dynamic scene reconstruction system using multiple commodity depth cameras that offers accurate surface model. Instead of expensive multiview scene capturing setups, our system only requires four Kinects which are located to cover the 360 degree surface of objects.We describe the capturing system setup, calibration method between cameras, and reconstruction processes including data acquisition, data integration, and visualization. We introduce a novel depth synthesis method for point cloud densification and noise removal in the depth data. Also, a new weighting function is presented to overcome drawback of existing volumetric representation method. In addition, we provide metrical evaluations and a computational analysis on various real-life datasets