GSA Annual Meeting in Indianapolis, Indiana, USA - 2018

Paper No. 173-1
Presentation Time: 9:00 AM-6:30 PM


TEWKSBURY, David A., Department of Geosciences, Hamilton College, 198 College Hill Rd, Clinton, NY 13323-1218

Our department, like many other geo departments, is keenly interested in the potential of Unmanned Aerial Systems (UASs) for collecting data for research purposes. Spatial accuracy is critical for many projects such as those involving mapping and change detection. This accuracy must be repeatable over multiple student projects and multiple flights occurring across hours to years. We have learned a number of lessons over the past two years as we have explored what can be achieved with our reasonably priced UAV, an unmodified Phantom 4 Pro.

Lesson 1: Understand what GPS data are collected by your UAV, what your processing software does with that data to generate the end products that you need, and the limits of spatial accuracy that can be achieved using the UAV alone. Our testing shows that a stock Phantom 4 Pro is capable of repeatable automated flights over the same area with X,Y offsets of 0.5 to 1 m between flights, depending on GPS satellite configuration at flight time. Z values, however, vary dramatically from flight to flight due to the fact that the Phantom 4 Pro calculates initial elevation based on barometric pressure at takeoff. Because barometric pressure varies, this calculation can result in Z value offsets of 10s of meters among multiple datasets of the same area. Data sets align closely in X and Y but are stacked vertically at different elevations.

Lesson 2: Use ground control points (GCPs) to improve spatial accuracy of derived products, and understand the method you use for determining X, Y, and Z coordinates of GCPs. We use 4’ X 4’ black and white aerial targets evenly distributed across the flight area. We collect coordinates with a Trimble Geo7X receiver with H-Star and correct through Trimble Pathfinder Office software, which queries local CORS stations. Once the imagery is loaded into SfM software, corrected GCP positions are tagged to corresponding target GCPs that appear in multiple images. Although placing GCPs and collecting position data can add 2 hours or more to a mission profile, we have found that decimeter-level accuracy in X, Y, and Z can be achieved, and generated DEMs and orthophotos align well with NYS orthoimagery. It is critical, however, to know whether Z values collected by your device are reported in height above mean sea level or height above the ellipsoid and, if the latter, what ellipsoid is being used.

  • GSA_2018_D_Tewksbury_opt.pdf (4.2 MB)