Toward faster 3D environment modeling based on feature extraction results from movies; Research overview and initiatives

Nakamura, Keita*; Hanari, Toshihide ; Madokoro, Hirokazu*; Imabuchi, Takashi  ; Kawabata, Kuniaki   ; Nix, S.*; Doi, Akio*

This paper introduces a research effort for faster 3D environment modeling of the workspace for decommissioning activities at the Fukushima Daiichi Nuclear Power Plant, applying as input video images shot during an investigation of the reactor containment vessel and reactor building. Especially, we construct a system that allows the selection of a 3D reconstruction method with as much information as possible within a specified time limit. We challenge this study with three methods: photogrammetry, simulation, and AI technology. Finally, we aim to integrate the results of each research to build a prototype system that automatically generates a more informative 3D reconstruction result within a specified time and according to the extracted feature values.