Abstract
Small uncrewed aerial systems (UASs) generate imagery that can provide detailed information regarding condition and change if the products are reproducible through time. Densified point clouds form the basic information for digital surface models and orthorectified mosaics, so variable dense point reconstruction will introduce uncertainty. Eucalyptus trees typically have sparse and discontinuous canopies with pendulous leaves that present a difficult target for photogrammetry software. We examine how spectral band, season, solar azimuth, elevation, and some processing settings impact completeness and reproducibility of dense point clouds for shrub swamp and Eucalyptus forest canopy. At the study site near solar noon, selecting near infrared camera increased projected tree canopy fourfold, and dense point features more than 2 m above ground were increased sixfold compared to red spectral bands. Near infrared (NIR) imagery improved projected and total dense features two- and threefold, respectively, compared to default green band imagery. The lowest solar elevation captured (25°) consistently improved canopy feature reconstruction in all spectral bands. Although low solar elevations are typically avoided for radiometric reasons, we demonstrate that these conditions improve the detection and reconstruction of complex tree canopy features in natural Eucalyptus forests. Combining imagery sets captured at different solar elevations improved the reproducibility of dense point clouds between seasons. Total dense point cloud features reconstructed were increased by almost 10 million points (20%) when imagery used was NIR combining solar noon and low solar elevation imagery. It is possible to use agricultural multispectral camera rigs to reconstruct Eucalyptus tree canopy and shrub swamp by combining imagery and selecting appropriate spectral bands for processing.
Highlights
Modern photogrammetric processing software, high-precision global navigation satellite systems (GNSSs), high-resolution consumer cameras, and mobile computing power have combined to place remote sensing imagery and photogrammetry products in the hands of field ecologists and environmental scientists [1,2]
The proportion of total reconstructed dense points representing trees (> 2 m above ground level (AGL)) was lowest in all spectral bands for the ST2 imagery set (5.8%–29.4%), which coincided with maximum solar elevation (Table 1)
We demonstrate for Eucalyptus canopy that both environmental and processing decisions inherently impact the final densified 3D models
Summary
High-precision global navigation satellite systems (GNSSs), high-resolution consumer cameras, and mobile computing power have combined to place remote sensing imagery and photogrammetry products in the hands of field ecologists and environmental scientists [1,2]. Routine monitoring requires that measurements are repeatable and reproducible through time for features of concern [6]. The reproducibility of equivalent products from independent imagery or software using automated photogrammetry processes is currently uncertain [7,8,9]. Tree canopy extent is a core monitoring requirement as the target communities are typically treeless [10] and currently indirectly mapped by the boundary of the surrounding forest canopy [10,11]
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have