Accuracy assessment of vertical and horizontal coordinates derived from Unmanned Aerial Vehicles over District Six in Cape Town

Abstract

Unmanned aerial vehicles (UAVs) are now an alternative for geospatial data collection for a variety of applications. One such application is terrain mapping, particularly digital elevation model (DEM) and digital surface model (DSM) products, but questions remain about its accuracy and efficiency, especially when compared to traditional ground survey methods. Thus, the purpose of this paper is to compare the traditional surveying methods for topographical mapping through datasets obtained through Total Station (Trimble M3) and a camera mounted on a quadcopter (DJI Phantom 4 Pro UAV). We obtained both datasets in the same location at the District 6 open field area in the City of Cape Town, with undulating terrain.  We compared the resultant horizontal coordinates (x and y) and orthometric height (H) at 159 check points (CPs). The drone-based elevations were derived using UAV drone computer vision techniques. In using the UAV drone, the reconstructed camera positions and terrain features were used to derive ultra-high-resolution point clouds, ortho-photos, and digital surface models from the multi-view UAV camera photos taken at 120 m above mean sea level. The root-mean-square-errors (RMSEs) for the differences between the Total Station and UAV coordinates at 159 CPs are ±0.046, ±0.038 and ±0.079 m for x, y, and H coordinates, respectively. Comparisons over slope ranges show that the highest orthometric height error (±0.090 m) is at a slope steepness of >15°, while the least orthometric height error (±0.073 m) is at a slope steepness of 5 - 10°. The results also show that the highest errors in x (±0.070 m) and y (±0.074 m) occur at a slope steepness of >15° and the least errors in x (±0.035 m) and y (±0.029 m) at a 5 – 10° slope steepness.  Similar comparisons on elevation ranges show that the highest orthometric height error (±0.098 m) is at an elevation range of 60 – 100 m, while the least orthometric height error (±0.052 m) is at a 40 – 60 m elevation range.  The highest errors in x (±0.051 m) and y (±0.061m) occur at elevation ranges of 40 – 60 m and 0 – 20 m, respectively, while the least errors in x (±0.040 m) and y (±0.025 m) occur at elevation ranges of 80 – 100 m and 40 – 60 m, respectively. These results indicate that horizontal positions (x, y) and orthometric heights (H) obtained from UAV are accurate enough for most mapping applications.