Topics

ACM MobiSys 2020にてBest Poster Awardを受賞しました
Received the Best Poster Award at ACM MobiSys 2020

以下の論文が、モバイルシステムのトップカンファレンスの1つであるACM MobiSys 2020 (The 18th ACM International Conference on Mobile Systems, Applications, and Services)にて、Best Poster Awardを受賞しました。
https://www.sigmobile.org/mobisys/2020/ (外部サイト)

AirPlanes: A DIY Modeling System Using Mobile VSLAM for Indoor Space
Taiga Nishiyama (Ritsumeikan University), Daichi Yoshikawa (Ritsumeikan University), Nobuhiko Nishio (Ritsumeikan University), Kota Tsubouchi

The following paper won the Best Poster Award at ACM MobiSys 2020 (The 18th ACM International Conference on Mobile Systems, Applications, and Services).
https://www.sigmobile.org/mobisys/2020/ (external link)

AirPlanes: A DIY Modeling System Using Mobile VSLAM for Indoor Space
Taiga Nishiyama (Ritsumeikan University), Daichi Yoshikawa (Ritsumeikan University), Nobuhiko Nishio (Ritsumeikan University), Kota Tsubouchi

Abstract:
3D space modeling is crucial for AR/VR applications and navigation services in indoor environments. However, the conventional approaches require specialized devices, such as cameras, depth sensors combined with structured-light/laser pointers, 3D laser scanners, and so on. The requirements of specialized devices are potential barriers to the popularization of 3D space modeling. This paper proposes a brand new "DIY" 3D space modeling system, called "AirPlanes". AirPlanes utilizes VSLAM technology, which has become available for mobile OS'es and generates planes in a 3D space model without the need for any proprietary sensors. The system uses a smartphone and a laser pointer and takes a laser-based active-sensing approach to plane-model generation integrated with VSLAM. The device is a “do-it-yourself” one combining a smartphone and off-the-shelf laser pointer. It estimates the projected laser dot's world coordinates by triangulation in real time. Through integration with VSLAM-based self-positioning, it makes 3D plane modeling possible, even for flat and monotone walls that current feature-point-based VSLAM recognition technology cannot measure.