Authors: Jonathon Chester*, Department of Ecosystem Science and Management, Pennsylvania State University, Andrew Yoder, Department of Geography, Pennsylvania State University, Philip Walsh, Department of Geography, Pennsylvania State University, Doug Miller, Department of Ecosystem Science and Management, Pennsylvania State University, Brennan Holderman, Department of Ecosystem Science and Management, Pennsylvania State University, Scott Drzyzga, Department of Geography-Earth Science, Shippensburg University of Pennsylvania , Brian Naberezny, Department of Civil and Environmental Engineering, Pennsylvania State University, Nooreen Meghani, Department of Geology, University of Illinois Urbana-Champaign, Jonathan Duncan, Department of Ecosystem Science and Management, Pennsylvania State University
Topics: UAS / UAV, Drones, Remote Sensing
Keywords: small unmanned aircraft systems, structure from motion, Photoscan, Pix4D, photogrammetry
Session Type: Poster
Start / End Time: 9:55 AM / 11:35 AM
Room: Lincoln 2, Marriott, Exhibition Level
Presentation File: No File Uploaded
Structure-from-motion (SfM) and multi-view-stereo (MVS) have become common photogrammetric techniques leveraged by earth scientists for acquiring high resolution orthomosaics, 2D digital elevation models, and 3D point clouds. Using non-metric cameras mounted on small unmanned aircraft systems (sUAS or “drones”), the SfM-MVS workflow offers a relatively low-cost approach for acquiring elevation data with high spatiotemporal resolution. Several software solutions, differing in price, available technical support, graphical user interface, and level of parameterization, currently implement a SfM-MVS workflow. These differences raise the fundamental question of how software selection influences drone-derived elevation data. The present study compared elevation products generated with two commercial SfM-MVS software packages. We used automated drones to capture nadir images at three study sites across a forest-wetland-urban land cover gradient, and processed the non-metric images using similar parameters in (1) Agisoft PhotoScan and (2) Pix4D Mapper. SfM models were georeferenced using dGNSS prior to applying MVS densification and point cloud interpolation. For each study site, we assessed 3D correspondence between dense clouds using Model-to-Model Cloud Comparison, and leveraged DEMs of Difference to evaluate 2D raster agreement. These data reveal general sub-decimeter elevation correspondence with highest overall agreement in the urban study site. In all three land covers, model correspondence was greatest on bare-earth surfaces, though it suffered at edges of vertical features due to differences in image alignment and processing. Results presented here provide a necessary insight into understanding the influence of software selection on elevation data derived using drone-based SfM-MVS photogrammetry across differing land covers.