A Review of Neural Radiance Field Approaches for Scene Reconstruction of Satellite Remote Sensing Imagery
High-resolution satellite remote sensing images have been recognized as an indispensable means for understanding geographical spaces,and their role in areas such as urban mapping,ecological monitoring,and navigation,has become increasingly important.The use of satellite remote sensing images for large-scale 3D reconstruction of the Earth's surface is currently a subject of active research in the fields of computer vision and photogrammetry.Neural Radiance Fields(NeRF),which utilizes differentiable rendering to learn implicit representations of scenes,has achieved the most realistic visual effects in novel view synthesis tasks of complex scenes and has attracted significant attention in the field of 3D scene reconstruction and rendering.Recent research has been primarily focused on using neural radiance field technology to extract scene representation and reconstruction from satellite remote sensing images.Ray space optimization,scene representation optimization,and efficient model training are mainly focused on by the neural radiance field methods for satellite remote sensing images.The latest progress in the application of neural radiance field technology in satellite remote sensing is comprehensively summarized in this paper.First,the basic concepts of neural radiance field technology and related datasets are introduced.Then a classification framework of neural radiance field methods for satellite remote sensing images is proposed to systematically review and organize the research progress of this technology in the field of satellite remote sensing.The relevant results of the application of neural radiance field technology in actual satellite remote sensing scenarios are detailed.Finally,analysis and discussion are conducted based on the problems and challenges faced by current research,and future development trends and research directions are prospected.
万方数据
维普
