Assessing the accuracy of unmanned aerial vehicles photogrammetric survey

Unmanned Aerial Vehicles (UAVs) have seen an increase in use on terrain mapping, especially in lieu of manual surveying methods. This study will evaluate the accuracy that can be achieved using a commercially available UAV and Structure-from-Motion (SfM) software. Photogrammetric data were captured with a UAV at three elevations three separate times. Sixteen aerial targets were located by a surveyor and used to geo-reference the models. Absolute accuracy was principally evaluated by measuring the distance between the computed location and the location determined by the surveyors. The study found that the most accurate survey conducted had a horizontal error of 0.88 cm (0.029 ft) and a vertical error of 0.38 cm (0.012 ft). The study also found that there was very little statistically significant difference in accuracy between surveys created with images captured from flights at the same altitude. An additional experiment conducted evaluated what the optimal number of ground control points was to yield the most accurate results. The study found that with a 7,500 m2 (82,500 sq. ft) site that there were significant improvements to the survey when three ground control points (GCPs) were included but only marginal improvements when four or more GCPs were added.