ふげん発電所における現場可視化システムの開発,2; 遺伝的アルゴリズムを用いたマーカ配置の最適化

Development of worksite visualization system for Fugen NPS, 2; Marker arrangement optimization using generic algorithm

石井 裕剛*; 下田 宏*; Bian, Z.*; 泉 正憲 ; 森下 喜嗣 

Ishii, Hirotake*; Shimoda, Hiroshi*; Bian, Z.*; Izumi, Masanori; Morishita, Yoshitsugu

拡張現実感を利用する際に必須の技術となるトラッキングの精度は、環境に貼り付けるマーカの配置によって大きく変化する。トラッキングを行う必要がある領域が複雑な場合や、マーカを貼り付けられる場所が制限される場合等には、人手により最適なマーカ配置を見つけることは困難である。本研究では遺伝的アルゴリズムを用いてマーカ配置の最適化を行う手法を開発し、その効果をシミュレーションにより評価した。

Improvement of tracking accuracy is an important issue when applying augmented reality to nuclear power plant fieldwork. Tracking accuracy depends highly on the marker arrangement when employing a tracking method using a camera and markers. Tracking accuracy becomes low if markers are pasted arbitrarily only on places that are easy to paste them onto. For those reasons, this study develops a marker arrangement optimization algorithm based on genetic algorithms. Tracking error caused by finite camera resolution is considered particularly in this study. A wheel tracking error computation method is developed to compute the tracking error from the marker arrangement. A genetic algorithm is adopted for obtaining higher tracking accuracy from an initial pool of marker arrangements using wheel tracking error computation as a fitness function. Trial results show that the tracking accuracy can be improved markedly by applying the marker arrangement optimization algorithm.