Asteroid Image Inpainting Method Based on Accelerated Diffusion Model

被引:0
作者
Zheng, Yunyun [1 ]
Huang, Xiangyu [1 ]
机构
[1] Beijing Inst Control Engn, Sci & Technol Space Intelligence Control Lab, Beijing, Peoples R China
来源
2024 9TH INTERNATIONAL CONFERENCE ON INTELLIGENT COMPUTING AND SIGNAL PROCESSING, ICSP | 2024年
关键词
diffusion model; image inpainting; asteroid impact; autonomous optical navigation;
D O I
10.1109/ICSP62122.2024.10743626
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
To mitigate the threat of asteroid impact on life on Earth, non-nuclear kinetic energy high-speed impact has been identified as the most viable strategy. To ensure accurate impact on the asteroid's center of mass with maximum kinetic energy, high-precision and high-speed autonomous optical navigation is necessary during the probe's approach to the asteroid. Autonomous optical navigation is crucial due to the difficulty of using historical tracking data such as astrometry on the ground, given the long distance between the asteroid and the probe. The navigation image is used to extract the centroid of the target asteroid, with the center of gravity of the asteroid profile serving as the impact point to provide a stable impact direction. It is important to note that asteroids have irregular shapes, fast spin, and rapidly changing bright regions. The shadow area of the asteroid is constantly changing due to illumination, which causes bias in the impact point due to the incomplete profile. To complete the contour of the target asteroid, a method based on the accelerated diffusion model (ADM) is proposed to inpainting the shadow area. ADM is fast enough for mission and has been found to have the highest accuracy index and the lowest centroid deviation after restoration when compared to several advanced image restoration methods.
引用
收藏
页码:604 / 607
页数:4
相关论文
共 12 条
[1]  
Dhariwal P, 2021, ADV NEUR IN, V34
[2]   Taming Transformers for High-Resolution Image Synthesis [J].
Esser, Patrick ;
Rombach, Robin ;
Ommer, Bjoern .
2021 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION, CVPR 2021, 2021, :12868-12878
[3]  
Goodfellow IJ, 2014, ADV NEUR IN, V27, P2672
[4]  
Ho J, 2020, P 34 INT C NEUR INF, P6840
[5]   Image-to-Image Translation with Conditional Adversarial Networks [J].
Isola, Phillip ;
Zhu, Jun-Yan ;
Zhou, Tinghui ;
Efros, Alexei A. .
30TH IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR 2017), 2017, :5967-5976
[6]  
[李建国 Li Jianguo], 2013, [中国空间科学技术, Chinese Space Science and Technology], V33, P16
[7]   RePaint: Inpainting using Denoising Diffusion Probabilistic Models [J].
Lugmayr, Andreas ;
Danelljan, Martin ;
Romero, Andres ;
Yu, Fisher ;
Timofte, Radu ;
Van Gool, Luc .
2022 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR), 2022, :11451-11461
[8]   Scalable Diffusion Models with Transformers [J].
Peebles, William ;
Xie, Saining .
2023 IEEE/CVF INTERNATIONAL CONFERENCE ON COMPUTER VISION, ICCV, 2023, :4172-4182
[9]  
Saharia C., 2022, ACM SIGGRAPH 2022 C, P1
[10]  
Song JM, 2022, Arxiv, DOI arXiv:2010.02502