Shadow Neural Radiance Fields for Multi-view Satellite Photogrammetry

被引:26
作者
Derksen, Dawa [1 ]
Izzo, Dario [1 ]
机构
[1] European Space Agcy ESTEC, Keplerlaan 1, NL-2201 AZ Noordwijk, Netherlands
来源
2021 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION WORKSHOPS, CVPRW 2021 | 2021年
关键词
3D; HEIGHT; STEREO;
D O I
10.1109/CVPRW53098.2021.00126
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
We present a new generic method for shadow-aware multi-view satellite photogrammetry of Earth Observation scenes. Our proposed method, the Shadow Neural Radiance Field (S-NeRF) follows recent advances in implicit volumetric representation learning. For each scene, we train S-NeRF using very high spatial resolution optical images taken from known viewing angles. The learning requires no labels or shape priors: it is self-supervised by an image reconstruction loss. To accommodate for changing light source conditions both from a directional light source (the Sun) and a diffuse light source (the sky), we extend the NeRF approach in two ways. First, direct illumination from the Sun is modeled via a local light source visibility field. Second, indirect illumination from a diffuse light source is learned as a non-local color field as a function of the position of the Sun. Quantitatively, the combination of these factors reduces the altitude and color errors in shaded areas, compared to NeRF. The S-NeRF methodology not only performs novel view synthesis and full 3D shape estimation, it also enables shadow detection, albedo synthesis, and transient object filtering, without any explicit shape supervision.
引用
收藏
页码:1152 / 1161
页数:10
相关论文
共 34 条
  • [1] [Anonymous], 2018, ARXIV180210249
  • [2] Bi S., 2020, EUROPEAN C COMPUTER, V2348, P294
  • [3] Bi Sai, 2020, Neural reflectance fields for appearance acquisition
  • [4] Automatic Large-Scale 3D Building Shape Refinement Using Conditional Generative Adversarial Networks
    Bittner, Ksenia
    Koerner, Marco
    [J]. PROCEEDINGS 2018 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION WORKSHOPS (CVPRW), 2018, : 1968 - 1970
  • [5] Bojanowski P., 2017, arXiv preprint arXiv:1707.05776
  • [6] Multitask Learning of Height and Semantics From Aerial Images
    Carvalho, Marcela
    Le Saux, Bertrand
    Trouve-Peloux, Pauline
    Champagnat, Frederic
    Almansa, Andres
    [J]. IEEE GEOSCIENCE AND REMOTE SENSING LETTERS, 2020, 17 (08) : 1391 - 1395
  • [7] De Franchis C., 2014, ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences
  • [8] In-Season Mapping of Irrigated Crops Using Landsat 8 and Sentinel-1 Time Series
    Demarez, Valerie
    Helen, Florian
    Marais-Sicre, Claire
    Baup, Frederic
    [J]. REMOTE SENSING, 2019, 11 (02)
  • [9] Automatic 3D Reconstruction from Multi-Date Satellite Images
    Facciolo, Gabriele
    de Franchis, Carlo
    Meinhardt-Llopis, Enric
    [J]. 2017 IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION WORKSHOPS (CVPRW), 2017, : 1542 - 1551
  • [10] The High Resolution Stereo Camera (HRSC) of Mars Express and its approach to science analysis and mapping for Mars and its satellites
    Gwinner, K.
    Jaumann, R.
    Hauber, E.
    Hoffmann, H.
    Heipke, C.
    Oberst, J.
    Neukum, G.
    Ansan, V.
    Bostelmann, J.
    Dumke, A.
    Elgner, S.
    Erkeling, G.
    Fueten, F.
    Hiesinger, H.
    Hoekzema, N. M.
    Kersten, E.
    Loizeau, D.
    Matz, K. -D.
    McGuire, P. C.
    Mertens, V.
    Michael, G.
    Pasewaldt, A.
    Pinet, P.
    Preusker, F.
    Reiss, D.
    Roatsch, T.
    Schmidt, R.
    Scholten, F.
    Spiegel, M.
    Stesky, R.
    Tirsch, D.
    van Gasselt, S.
    Walter, S.
    Waehlisch, M.
    Willner, K.
    [J]. PLANETARY AND SPACE SCIENCE, 2016, 126 : 93 - 138