Modeling forest canopy surface retrievals using very high-resolution spaceborne stereogrammetry: (II) optimizing acquisition configurations
Year: 2023
Type: article
Abstract: The optimal configurations for very high-resolution (VHR, <2 m) spaceborne imagery collection to support stereogrammetry over complex forested terrain remain uncertain. We conducted a comprehensive sensitivity study of digital surface models (DSMs) derived from thousands of simulated along-track VHR stereopairs over two lidar-reconstructed forested scenes of closed and open canopies using the discrete anisotropic radiative transfer (DART) model. We evaluated the influence of convergence angle (CA), solar illumination, and image resolution on the derived DSM accuracy relative to the reference DSM and digital terrain model (DTM) products from airborne lidar data. Our results confirmed that the CA is the most critical acquisition parameter for DSM accuracy. Compared to the frequently used CA of ∼35∘ for along-track stereopair acquisitions by WorldView satellites, a smaller CA can provide better accuracy for forest canopy shape estimation by reducing occlusions and mitigating radiometric variance caused by the bidirectional reflectance characteristics of vegetation. For forested scenes over relatively flat terrain, oblique solar zenith angles (50−70∘) yielded more consistent DSMs with better accuracy, whereas images with a hotspot configuration generated elevations that were closer to the DTM. Image pairs with smaller ground sample distance (GSD) improved the DSM accuracy, and combinations of small (nadir) and large (off-nadir) GSDs had accuracy between those derived from homogeneous GSDs. These simulation results suggest that available global archives of DSMs from VHR stereo imagery collected under a range of acquisition configurations will yield inconsistent estimates of canopy surfaces. This study also provides a benchmark dataset and configuration guide for 1) selecting existing data to retrieve the forest canopy surface shape, and 2) defining requirements for future satellite missions to characterize the forest canopy surface using stereogrammetry.
Source: Remote Sensing of Environment
Authors: Tiangang Yin(Conceptualization, Investigation, Methodology, Software, Validation, Writing – original draft, Writing – review & editing), Paul M. Montesano(Data curation, Investigation, Methodology, Software, Writing – original draft), Bruce D. Cook(Data curation, Investigation, Methodology, Supervision, Writing – review & editing), Eric Chavanon(Methodology, Software, Validation, Visualization), Christopher S.R. Neigh(Conceptualization, Methodology, Software, Supervision, Writing – review & editing), David Shean(Conceptualization, Methodology, Software, Writing – review & editing), Dongju Peng(Investigation, Methodology, Software), Nicolas Lauret(Investigation, Methodology, Software), Ameni Mkaouar(Investigation, Validation, Writing – review & editing), Omar Regaieg(Investigation, Methodology, Validation), Zhijun Zhen(Investigation, Methodology, Validation), Rongjun Qin(Software, Validation, Writing – review & editing), Jean-Philippe Gastellu-Etchegorry(Conceptualization, Investigation, Software, Supervision), Douglas C. Morton(Conceptualization, Investigation, Supervision, Writing – original draft, Writing – review & editing)
Institutions: Goddard Space Flight Center, Hong Kong Polytechnic University, Earth System Science Interdisciplinary Center, University of Maryland, College Park, Science Systems and Applications (United States), Centre National de la Recherche Scientifique, Centre National d'Études Spatiales, Centre d'Études Spatiales de la Biosphère, Institut de Recherche pour le Développement, Université de Toulouse, University of Washington, Nanyang Technological University, Earth Observatory of Singapore, University of Maryland, Baltimore County, The Ohio State University
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
Estimated APC paid: $4,070
OpenAlex ID: W4387071022
