In this paper, we sought to determine the accuracy of UAV photogrammetry relative to GNSS RTK survey for stockpile surveys, at an average photo scale of 1:6 600. Vertical accuracy of the UAV photogrammetry with respect to the GNSS RTK sur- vey (combining the results from Tables 2 and 4) was found to be RSMEz = 0.044 m. According to the US National Map Accuracy Standards (NMAS), the equivalent contour interval would be 0.145 m [SBCD 1998]. Horizontal accuracy of the UAV aer- ial mapping with respect to the GNSS RTK survey (Table 4) was found to be RSMEr = 0.039 m. According to ASPRS standards, this is sufficient for mapping at a scale of 1:200. Traditionally, this scale was considered to be beyond the practical limit of aerial mapping, being more appropriate for ground survey methods [ASPRS 1988].
Volumes were computed for the GNSS surface and the two UAV photogrammetric surfaces. The UAV surface agreed with the GNSS surface to 0.7%, and the two UAV surfaces agreed within 0.2% of total pile volume (530 255 m3). It is inter- esting to note that the volume difference between the two UAV surfaces only amounted to a 1-cm dif- ference over the surface area of the pile–an almost trivial amount given the accuracies observed.
We also looked at the repeatability of UAV photogrammetric surveys by comparing DSMs from two flights with different ground control. The mean absolute C2C distance between the two DSMs was 0.032 m ± 0.024 m (for 9.8 million points). This indicates that at a 95% confidence level, the two DSMs can be considered the same surface.
The results of this paper support the findings of others with respect to the accuracy and mapping suit- ability of UAV photogrammetry [Draeyer et al. 2014; Savoy 2013; Strecha 2014; Wang et al. 2014]. The accuracies achieved are not the highest the in- house developed UAV system described in this paper should be able to attain. The described UAV system [Mah 2014] can easily fly at half the height above ground observed in this paper and still maintain 75% forward and side overlap in images. It is not unrea- sonable to estimate that accuracies could approach half of those stated above. At this level, extremely accurate ground control would be required.
It is our opinion that the results support the use of UAV aerial photography for the accurate survey of stockpiles and earthworks. The UAV aerial pho- togrammetric approach is accurate, efficient, and can provide increased safety over conventional terrestrial survey methods.