Recently, our partner Luke Wijnberg with 3Dronemapping showed off one of our E384 UAVs to a surveying company in Free State, South Africa. The surveying company compared the pictures taken from our plane with those from a competing company, and our pictures were much sharper and showed perfect alignment. Below is a summary Luke wrote about his experience with the surveying company, including the ins and outs of the flight and how it compared to the competition.
Aerial Survey of N5(Bethlehem-Wilge River), Harrismith, Free State
A demonstration survey was requested by a surveying company to determine whether or not the 3Dronemapping E384 fixed wing was a suitable for their common survey applications. The demonstration location was withheld from 3Dronemapping until we arrived on at their company offices.
The surveying company placed 9 GCP in various styles across the site based on the NRB system. These were surveyed using a rapid – static GPS system and claimed a RMS of 25mm YXZ. It should be noted that the control was unverified by 3DM and a scaling error was perhaps present. The distribution of control was uneven and denser toward the N of the site. The control was not clearly defined on the ground and was often placed near to tall trees.
On 9 December 2014, LW travelled to Harrismith, Free State to meet with the surveying company for purposes of demonstrating the E384. At their offices, the area proposed for aerial survey was made clear and basic mission planning began. The company requested that the survey be done at 70m AGL and gave an LZ. 3DM suggested 60% sidelap and 78% overlap as a strong 9m/s NW was blowing and would allow for some redundancy with image movement. At the LZ, it was suggested by 3DM to move to another location 500m E to accommodate for the NW wind and avoid the dense network of powerlines.
At the LZ, the E384 was prepared for flight and the mission waypoints issued to the aircraft. KB contacted the local ATC (11km N from the LZ) to request permission to fly and advisory. The flight took approximately 28 minutes, 22.93km flown and belly landing occurred 50m SE of LZ with no input necessary.
The images were processed using Agisoft Photoscan Pro. Every image was examined for clarity and banking images were discarded. A batch process was setup using the following settings:
Alignment(Medium, pair selection-Generic)
Dense Cloud generation(Lowest)
Build mesh(Sparse Cloud)
Build Texture(Adaptive Orthophoto)
After 1.2hours on a I7, 16gig memory, 4gb video card, a basic 3d model was formed.
Pre surveyed control was inserted into APP and verified. After optimising the point cloud, the model was regenerated using a higher accuracy of precision and a denser cloud developed. A final orthoimage and DTM was generated at 0.02m/pixel and 0.15m respectively.
During flight planning, it was revealed that a competing UAV company had 8 days previously done a demo over the same site using a fixed wing craft and same camera. The results were issued to me and a comparison formed. The E384 produced much clearer results than the competitor, Ghosting and image blur was found throughout their ortho. It is suggested that APP was also used to generate their model and the same control used. LW is of the opinion that insufficient overlap was the culprit as some images are clear(not as clear as E384) and that the image needed to be extrapolated to accommodate the lack of stereo pair. Other suggestions are that their lense was dirty, UAV was flying too fast, unstable platform unsuitable for aerial mapping. On the competitor’s flight, no wind was reported and clear cloudless conditions experienced. The DTM was compared against traditional ground survey and discrepancies were found 0.03m. This is well within “Class A”. The DTM from the competitor was compared and no significant differences found.
The surveying company realised that importance of clear, well distributed control, was essential to producing accurate 3D models
The Competitor –E384 comparison shows that expensive does not necessarily mean better quality. The surveying company also chose 3DM due to a willingness to offer assistance and guidance with this new type of survey due to an understanding of the limitations and technical application of this low level photogrammetry as well as their experience in this field.
Flight path 70m AGL
Competitor image on the left showing blurring. E384 on the right is much crisper and sharper.
Competitor image on the left showing ghosting. E384 on the right has none and perfect stereo alignment.
Competitor image on the left. E384 on the right showing high level of detail due to stabilised flight.
Want to stay up to date on our latest findings? Sign up for our newsletter!