Abstract
The Ice, Cloud, and Land Elevation Satellite-2 (ICESat-2) photon data is the emerging satellite-based LiDAR data, widely used in surveying and mapping due to its small photometric spot and high density. Since ICESat-2 data collect weak signals, it is difficult to denoise in shallow sea island areas, and the quality of the denoising method will directly affect the precision of bathymetry. This paper proposes a back propagation (BP) neural network-based denoising algorithm for the data characteristics of shallow island reef areas. First, a horizontal elliptical search area is constructed for the photons in the dataset. Suitable feature values are selected in the search area to train the BP neural network. Finally, data with a geographic location far apart, including daily and nightly data, are selected respectively for experiments to test the generality of the network. By comparing the results with the confidence labels provided in the official documents of the ATL03 dataset, the DBSCAN algorithm, and the manual visual interpretation, it is proved that the denoising algorithm proposed in this paper has a better processing effect in shallow island areas.
© 2022 Optica Publishing Group
Full Article | PDF ArticleMore Like This
Guoan Huang, Zhipeng Dong, Yanxiong Liu, Yilan Chen, Jie Li, Yanhong Wang, and Wenjun Meng
Opt. Express 31(25) 41496-41517 (2023)
Xiaomei Lu, Yongxiang Hu, Ali Omar, Yuekui Yang, Mark Vaughan, Sharon Rodier, Anne Garnier, Robert Ryan, Brian Getzewich, and Charles Trepte
Opt. Express 30(20) 36509-36525 (2022)
Yiming Song, Yuli Han, Zhaowang Su, Chong Chen, Dongsong Sun, Tingdi Chen, and Xianghui Xue
Appl. Opt. 63(1) 275-282 (2024)