25. November 2024, 15:00 until 15:30

Master defense Markus Brezovsky

Other

Comparison of 3D reconstructions using Neural Radiance Fields and Dense Image Matching

With advances in photogrammetry and computer vision, Neural Radiance Fields (NeRFs) have emerged as an innovative method for 3D reconstruction. This paper investigates the performance of NeRFs compared to traditional Dense Image Matching (DIM) methods using data captured with a DJI M350 drone and a Zenmuse P1 camera. A terrestrial laser scanner point cloud serves as a reference. Two different objects - a group of trees and a chapel - were analyzed under varying flight paths (linear and orbital) to evaluate the influence of the recording geometry on the reconstruction quality.The analysis includes quantitative and qualitative comparisons of the point clouds as well as a discussion on the practicality of NeRFs in UAV mapping. Finally, the challenges and potential of this new technology for future applications in geodesy and geoinformation are outlined.

Calendar entry

Event details

Event location
FH HS 7, 2nd floor yellow
1040 Wien
Wiedner Hauptstraße 8
Organiser
TU Wien
Public
Yes
Entrance fee
No
Registration required
No