Extracting locations from individual cameras: Single view three-dimensional reconstruction to extract flow velocities from static cameras

Stuart Mead^1, Gert Lube¹, Ermanno Brosch¹

• Affiliations: ¹Volcanic Risk Solutions, Massey University, New Zealand

• Presentation type: Poster

• Presentation time: Monday 16:30 - 18:30, Room Poster Hall

• Poster Board Number: 138

• Programme No: 3.1.30

• Theme 3 > Session 1

---

Abstract

The estimation of flow velocities for volcanic phenomena can provide useful information for further research into flow behaviour and reconstruction of event timelines. However, this estimation can be difficult without a-priori knowledge of an event and noise or availability affects results from some sensors (e.g. infrasound arrays, seismometers). In the post-event analysis of the 9 December 2019 Whakaari pyroclastic surges, we were able to estimate flow velocities by reprojecting webcam images into geographic coordinates using a single webcam and computer vision techniques. The reprojection of single camera images into world coordinates, called 'single-view reconstruction', is vastly more difficult than multi-view (e.g. structure-from-motion) reconstruction. However, provided some information on camera location and geometry of the terrain, we found it was possible in our case through solution of the Perspective-n-Point problem. This technique may be applicable to other events with static webcams that are prevalent on many volcanoes, provided appropriate information and camera positioning are available. We demonstrate this through analysis of the Whakaari example, evaluating the accuracy of our solution, provide the code and discuss limitations and constraints to this single view reconstruction.