14 June 2017 ... camera and an Emlid Reach for PPK work. ... The same images were also geotagged in PPK fashion from the Reach and its camera event sync.
Investigating the geometric accuracy of Propeller AeroPoints and Propeller cloud processing Report compiled by Luke Wijnberg and Jonathan Swart from 3DroneMapping 14 June 2017
Contents
1. Introduction 2. Methodology 3. Results 3.1 Recorded XYZ data 3.2 Cross sections 4. Final thoughts
1.
Introduction
The Propeller AeroPoint is a smart ground control point (GCP) which aims to make the gathering of accurate data simple and affordable during drone surveys. These lightweight and durable units repeatedly record positional data during UAV flights, with each AeroPoint working collaboratively against each other; therefore the more AeroPoints used, the greater the accuracy of each unit.
Figure 1:
The Propeller AeroPoint
Propeller claims that their AeroPoints will work anywhere in the world with geometric accuracies of 20mm (X), 20mm (Y) and 50mm (Z). This statement was put to the test in Cato Ridge, South Africa, with a pre-planned AOI that would be the testing ground for various geometric comparisons to be conducted against 3DroneMapping's (3DDM) tried, tested and proven drone surveying methods.
Figure 2:
Area of interest - Cato Ridge, South Africa [Image adapted from: Google 2017].
2.
Methodology
To successfully investigate the geometric accuracy of the AeroPoint technology, two tests were conducted in order to analyse and compare the resulting data. The two tests were as follows: Test 1 - AeroPoints versus traditional survey-grade GPS control measurements 10 AeroPoints were roughly distributed around the perimeter of the site and were activated to log raw GPS data. The exact location of the AeroPoints were measured using a survey grade Emlid Reach RS GPS in RTK mode as well as logging raw data for post processing. Some points were purposely placed in technically challenging areas such as under power lines in order to test the robustness of the final coordinate. After two hours of data collection, the AeroPoints were removed and the raw data automatically uploaded to the Propeller processing servers. The coordinate for the “LZ” was determined prior to the exercise and was issued to Propeller to coordinate or reference itself too. This was done in order to find commonality between the survey tests. A Reach RS was set with a 2m vertical offset over “LZ”. This was set to broadcast raw RTCM over a LoRa radio to a rover. The base also recorded raw measurements at 1HZ for later post-processing. The RTK results were exported and compared to the results issued from Propeller. Both sets of coordinates were transformed from WGS84 (EPSG:4326) to Hartebeeshoek Lo31 (EPSG:2054) for ease of coordinate comparisons. Test 2 - Propeller Platform processed data versus locally processed PPK survey 843 images were collected from a fixed wing platform. On-board sensors included a high resolution camera and an Emlid Reach for PPK work. As a test for the AeroPoints and their performance with the Propeller cloud photogrammetry processing software, images were geotagged using only the uncorrected navigational GPS (+-5m accuracy). The 8.29GB of data was uploaded to Propeller and set to process; no further options were made available. The resultant orthophotos and DEM were downloaded, re-projected to Hartebeeshoek Lo31 (EPSG:2054) and compressed for later comparison. The same images were also geotagged in PPK fashion from the Reach and its camera event sync system. The events and track-log were further post processed using “LZ” as a reference or base for the survey. A Reach RS was set with a 2m vertical offset over “LZ” and recorded raw measurements at 1HZ for post processing. The on-board Reach GPS recorded raw GPS data at 14HZ. Post processing was done in RTKLIB V2.4.3 b27 with 100% of the events being resolved to 1cm accuracy. Photogrammetry was undertaken locally with Pix4D (V3.2) with appropriate settings for high accuracy geotagging. No additional ground control was used to reference the model. Exported othophotos and DTM were compressed and used in the final comparisons. 3.
Results
Following a successful day in the field, both parties began processing their gathered data and running the necessary tasks using the preferred specialised software. The results provided informative data from the different systems used and were ideal for investigating the accuracy of the AeroPoints.
Figure 3:
3DDM orthophoto (0.03cm) and location of survey points
3.1 BASE LZ PP1 PP2 PP3 PP4 PP5 PP6 PP7 PP8 PP9 Q1 Q2 Q3 TR15
Table 1:
Comparisons of recorded XYZ data (meters)
(Control)ReachRS X Y Z 34008.710 3286059.367 789.923 34010.128 3286061.944 788.130 34139.037 3286208.770 788.550 33763.654 3286391.681 788.490 33694.051 3286607.410 790.425 33988.931 3286673.423 792.907 34181.620 3286577.671 791.045 34412.123 3286395.020 792.045 34648.182 3286185.562 791.125 34467.504 3286142.310 794.345 34261.193 3286084.558 791.455 34382.684 3286468.532 791.785 34648.168 3286215.805 791.805 34270.458 3286062.010 791.245 33763.725 3286441.083 792.300
PPK(Digitised) X Y LZ PP1 PP2 PP3 PP4 PP5 PP6 PP7 PP8 PP9 PKD1 PKD2 PKD3 PKD4 PKD5 PKD6 PKD7 PKD8 PKD9
34010.118 34139.061 33763.704 33694.056 33988.931 34181.663 34412.101 34648.149 34467.496 34261.173 33706.042 33839.391 34010.612 34296.350 34429.692 33882.605 33920.002 33856.305 34239.073
3286061.944 3286208.780 3286391.644 3286607.444 3286673.463 3286577.669 3286395.016 3286185.534 3286142.352 3286084.571 3286637.837 3286667.863 3286411.781 3286529.630 3286235.628 3285942.650 3286257.084 3286518.947 3286303.998
GPS Control to PPK dX dY dZ
Z 788.117 788.532 788.498 790.454 792.927 791.040 792.049 791.086 794.296 791.435 790.220 791.532 791.606 792.495 795.543 784.710 788.778 790.946 790.256
-0.01 0.02 0.05 0.01 0.00 0.04 -0.02 -0.03 -0.01 -0.02
0.00 0.01 -0.04 0.03 0.04 0.00 0.00 -0.03 0.04 0.01
-0.01 -0.02 0.01 0.03 0.02 0.00 0.00 -0.04 -0.05 -0.02
3DDM's Control GPS and PPK
The above table compares 3DDM's control GPS which was collected using a ReachRS device, against digitised PPK data (orthophoto and DEM). This process was included as it shows a low level of variance of 0.001m between the GPS and PPK systems, proving that the GPS data is a true reference to use for determining the accuracy of the AeroPoint technology.
Aeropoints X Y LZ PP1 PP2 PP3 PP4 PP5 PP6 PP7 PP8 PP9 Table 2:
34010.128 34139.035 33763.636 33693.980 33988.859 34181.602 34412.111 34648.170 34467.528 34261.178
Z
3286061.944 3286208.764 3286391.692 3286607.530 3286673.519 3286577.685 3286395.031 3286185.523 3286142.320 3286084.548
788.130 788.570 788.531 790.496 792.999 791.066 792.094 791.138 794.398 791.458
GPS Control to Aeropoints dX dY dZ 0.00 0.00 0.02 0.07 0.07 0.02 0.01 0.01 -0.02 0.01
0.00 0.01 -0.01 -0.12 -0.10 -0.01 -0.01 0.04 -0.01 0.01
0.00 -0.02 -0.04 -0.07 -0.09 -0.02 -0.05 -0.01 -0.05 0.00
Comparing AeroPoints to 3DDM's GPS
The above table was part of the first conducted test and contains the XYZ raw data captured by the AeroPoints. This raw data was compared to the findings of 3DDM's ReachRS GPS (which have been proved to be extremely accurate in table 1). The comparisons drawn here have concluded excellent results between the two systems, with a calculated variance of 0.002m. Propeller Processed X Y LZ PP1 PP2 PP3 PP4 PP5 PP6 PP7 PP8 PP9 PD1 PD2 PD3 PD4 PD5 PD6 PD7 PD8 PD9
Table 3:
34010.115 34139.029 33763.641 33693.984 33988.855 34181.622 34412.104 34648.181 34467.518 34261.166 33706.027 33839.353 34010.585 34296.286 34429.673 33882.585 33919.992 33856.255 34239.007
3286061.980 3286208.792 3286391.692 3286607.540 3286673.484 3286577.718 3286395.039 3286185.511 3286142.344 3286084.536 3286637.979 3286668.022 3286411.845 3286529.674 3286235.611 3285942.590 3286257.133 3286519.043 3286304.017
Z 788.124 788.557 788.595 790.385 792.978 791.039 792.121 791.088 794.383 791.417 790.170 791.499 791.588 792.457 795.614 784.755 788.843 790.943 790.265
Propeller Processed to PPK dX dY dZ 0.00 -0.03 -0.06 -0.07 -0.08 -0.04 0.00 0.03 0.02 -0.01 -0.01 -0.04 -0.03 -0.06 -0.02 -0.02 -0.01 -0.05 -0.07
0.04 0.01 0.05 0.10 0.02 0.05 0.02 -0.02 -0.01 -0.03 0.14 0.16 0.06 0.04 -0.02 -0.06 0.05 0.10 0.02
0.01 0.02 0.10 -0.07 0.05 0.00 0.07 0.00 0.09 -0.02 -0.05 -0.03 -0.02 -0.04 0.07 0.04 0.07 0.00 0.01
Comparing the AeroPoint processed data to 3DDM's PPK system
For the second test, the processed AeroPoint data (using Propeller's own software) has been compared to the PPK digitized survey by 3DDM. Once again, the AeroPoints have gathered accurate results as this processed data has resulted in a comparison variance of 0.003m.
3.2
Cross sections (meters)
The variance of the recorded comparisons from the three tables above can also be observed in the form of cross sections. Cross sections were drawn in five different locations within the AOI over sections of different surface characteristics (ie. a concrete parking area, rough field, etc). Sections such as tarred roads and concrete parking zones are usually extremely flat and shouldn't show much variance between the different systems, however uneven surfaces such as within the rough field, could provide a harder test for accurate readings. GPS (red) against PPK (blue)
Figure 4:
Road near abattoir
Figure 5:
Small rubbish dump near the abattoir road
Figure 6:
Rough Field
Figure 7:
Parking area
Figure 8:
Semi-constructed house
GPS (red) against the AeroPoint's processed data (green)
Figure 9:
Figure 10:
Road near abattoir
Small rubbish dump near the abattoir road
Figure 11:
Rough field
Figure 12:
Parking area
Figure 13:
Semi-constructed house
4. Final thoughts Using two different methods to determine geometric accuracy, there is sufficient evidence to conclude that the Propeller AeroPoints and accompanying cloud processing are truly capable of delivering consistent, high-quality data. Although the AeroPoints are not as accurate as the more traditional surveying methods such as a PPK system, the added benefits of quick to deploy technology which is durable and user-friendly, certainly compensates for the small amount of geometric variance. Propeller's claim regarding their AeroPoints recording accurate data in any region of the world has also been confirmed in this case, as Australian designed GCP technology has captured admirable results in the Southernmost tip of Africa. However, with high geometric accuracy and such an easy to use solution for drone survey GCPs, one can expect costs to be high on acquiring such technology (6000USD for x10 AeroPoints, compared to 1600USD for 2 ReachRS RTK/PostProcessing units capable of placing unlimited GCPs). With initial high costing overlooked, Propeller’s AeroPoints can ultimately save drone survey companies time and money while not having to be concerned with negatively affected geometric accuracy.
