Date of Graduation
Master of Natural and Applied Science in Geography, Geology, and Planning
Geography, Geology, and Planning
photogrammetry, sUAS, UAV, DEM, orthophoto, gradual selection, point cloud filtering, SfM, mission design
Geographic Information Sciences | Physical and Environmental Geography | Remote Sensing | Spatial Science
This research investigates the influence that various flight plan and mission design strategies for collecting small unmanned aerial system (sUAS) imagery have on the accuracy of the resulting three-dimensional models to find an optimal method to achieve a result. This research also explores the effect that using gradual selection to reduce the sparse point cloud has on product accuracy and processing details. Imagery was collected in the spring of 2018 during leaf-off conditions at six field sites along the North Fork of the White River. The aerial imagery was collected using a DJI Phantom Pro 4 sUAS. Four different image acquisition missions were flown at each of the sites. Each of the base mission imagery sets were processed individually and in various combinations. The commercial Structure-from-Motion (SfM) photogrammetry software known as Agisoft PhotoScan was used to process the data and generate the Digital Elevation Models (DEMs) and orthophotos. Due to the high number of processing iterations required in this research, a script was developed to automate the point cloud filtering gradual selection process. Profile views were used to assess the differences between each mission design and to visualize systematic errors. In this investigation, the imagery set which consistently performed with high relative accuracy and low relative processing times was the NS Oblique imagery set utilizing automated gradual selection. Imagery sets created by combining two or more of the base mission photosets generally produced results with accuracy levels similar to or worse than the results of the NS Oblique imagery set and the other base mission imagery sets. Results produced with and without gradual selection were similar in most cases, however, gradual selection reduced dense cloud processing time by an average of 37%.
© Daniel Shay Hostens
Hostens, Daniel Shay, "Determining the Effect of Mission Design and Point Cloud Filtering on the Quality and Accuracy of SfM Photogrammetric Products Derived from sUAS Imagery" (2019). MSU Graduate Theses. 3372.