[A-CG47] Promotion of climate and earth system sciences using manned/unmanned aircrafts

[ACG47-02] Novel view synthesis of Typhoon Mindulle from images recorded in T-PARCII aircraft reconnaissance using neural radiance fields

*Takuya Mori¹, Yoshitaka Adachi¹, Eiji Kitajima¹, Hiroyuki Yamada¹, Kosuke Ito¹, Ryota Miyata¹ (1.University of the Ryukyus)

Keywords:Aircraft reconnaissance, Artificial intelligence, Typhoon

We synthesized several views around the eye of Typhoon Mindulle in 2021 based on the images taken through the upper-tropospheric reconnaissance flight (T-PARCII aircraft reconnaissance), using a machine-learning algorithm called neural radiance fields (NeRF).
Numerical calculations confirmed that NeRF could reproduce the appearance of clouds around the typhoon eye from not less than 30 images.
As the number of training images increased, NeRF became able to reproduce even the detailed could streaks (Figure 1).