AirInspect: Low-Cost Indoor 3D Reconstruction and Targeted Web Inspection with Smartphone-Mounted UAVs

Nojin Park1, Ji Hyun Seo1,2, Byounhyun Yoo1,3
1Intelligence and Interaction Research Center, Korea Institute of Science and Technology 2Department of Computer Science and Engineering, Korea University 3AI-Robotics, KIST School, Korea National University of Science and Technology
Video Code

AirInspect Low-Cost Indoor 3D Reconstruction and Targeted Web Inspection with Smartphone-Mounted UAVs

Abstract

AirInspect is a low-cost 3D reconstruction and remote inspection system built from commercial UAVs and consumer smartphones (e.g., iPhones). By relying on off-the-shelf hardware and existing software libraries, the system remains broadly compatible across different UAV vendors with minimal customization.

In our workflow, a drone-mounted smartphone performs autonomous indoor scanning while a server streams results to the web for real-time, multi-user viewing via A-Frame. The system first produces a coarse 3D reconstruction to provide users a global overview of the space. For practical inspection tasks, users can then request targeted remote monitoring to quickly examine specific regions of interest—such as cracks or defects—through high-resolution 2D imagery before triggering more time-consuming accurate reconstruction.

This dual-stage design emphasizes practicality, cost efficiency, and universality, making AirInspect a promising solution for real-world indoor inspection and monitoring scenarios.

iPhone Mount

iPhone Mount Image 1 (assembled view)
Assembled iPhone mount. Blue parts are rubber damping pads for vibration isolation.
iPhone Mount Image 2 (exploded view)
Exploded view showing the layered structure and blue rubber pads at the frame interface and phone grips.
iPhone Mount Image 3
iPhone Mount Image 4

iPhone Mount Design

We designed a lightweight iPhone mount for UAVs with three goals: vibration isolation, unobstructed camera FoV, and preserved center of gravity. Rubber damping pads (shown in blue in the CAD renders) are placed between the mount and the airframe and on the phone grips to decouple propeller-induced vibrations from the camera and IMU, improving real-time localization and 3D reconstruction. The mount keeps the phone offset from frame elements to avoid occlusions, and its position is tuned to align the combined mass with the vehicle’s centerline. The parts are 3D-printed in PLA (~50 g) and tested in flight, showing stable fixation and reduced drift error.

  • Vibration isolation: Blue rubber pads at frame interface and phone grips attenuate high-frequency jitter.
  • Camera visibility: Offset/angle placement prevents frame occlusion in the camera’s field of view.
  • Balance: Mount location preserves CG near the airframe axis for stable flight.
  • Build: PLA 3D-printed (~50 g), validated in flight with improved localization stability.

News

Teaser video uploaded — Short overview on YouTube. Open video.

iPhone Mount uploaded — Added assembly and exploded views in the iPhone Mount section with vibration-damping details.

Project website launched — Initial public page for AirInspect is live. Watch the teaser and check the iPhone Mount preview.

Contact

AirInspect is a open-source project from the Webizing Research Lab (WRL) at KIST, supervised by Prof. Byounghyun Yoo.

We study autonomous indoor scanning with smartphone-mounted UAVs, practical 3D reconstruction pipelines, and web-based multi-user inspection. It is maintained by the WRL team, see the team member list.

We welcome suggestions and collaboration from academia and industry—for example integration with facility workflows, reproducible datasets, deployment pilots, or media inquiries. If you are interested in using or contributing to AirInspect, please contact us.

BibTeX

@article{park2021nerfies,
  author    = {Park, Keunhong and Sinha, Utkarsh and Barron, Jonathan T. and Bouaziz, Sofien and Goldman, Dan B and Seitz, Steven M. and Martin-Brualla, Ricardo},
  title     = {Nerfies: Deformable Neural Radiance Fields},
  journal   = {ICCV},
  year      = {2021},
}