DDC20 Establishing Accuracy Q&A Responses

Good afternoon All! We didn’t have near enough time for Q&A after our session yesterday so I wanted to post some responses to the questions here, especially the ones we couldn’t get to. As a community let’s pick these apart. If we get some strings going I will split them out into their own threads.

Masih - What equipment do you use for GCPs? What are some alternates to Propeller GCPs?
We use Emlid RS2 Multi-Channel Receivers and Microsurvey FieldGenius. FieldGenius is only required if you need to localize to a jobsite using surface coordinates and you want to align with a CAD file. Otherwise for just map to map relative accuracy you can just use the Reachview software that comes with the Emlid kit. If you don’t necessarily need centimeter-level accuracy there are other less expensive options that will provide sub-meter or even decimeter accuracy.

Max - How do you validate your measurements? Do you do ground truth measurements regularly?
We utilize the checkpoint option in DroneDeploy’s processing and also import other survey observations into our Civil CAD software. There is an easy function in Carlson Civil that allows for a cut/fill report on a batch of points versus a surface. Those values can also be exhibited on a map.

Craig - Is the Phantom 4 Pro v2 the industry standard?
While I have no data to back it up I am pretty sure that the Phantom 4 series drones are the most common mapping drones as of right now. I am sure the Mavic 2 Pro put a dent in the foothold that the P4P had, but for the simple fact that the P4P models have a mechanical global shutter they will maintain their standing for some time or at least until they aren’t available (again) and we are forced to move on. There is also allot of interest in the Autel Evo II Pro and the Skydio 2. The Skydio is especially interesting to those who may want to work in confined space or indoors due to the fact that is has 360 awareness and some spatial AI.

Jack - What influences the amount of GCPs you decide to use during a survey?
First thing is whether or not you are using a PPK or RTK solution. This allows many less GCPs than just relying upon the onboard drone GNSS module. On average we would use 15-20 GCP’s on a 50 acre jobsite depending on its shape and the number continues to grow as the sites get larger somewhere in the order of 1 per every 10 additional acres. With PPK/RTK you can use anywhere from 5-7 on average with an additional 1 per every 100 acres.

Ben - Did you find it difficult to make the transition from just measuring stockpiles to mapping entire sites with GCP’s for cut / fill analysis?
I won’t say it is easy just for the fact of how much more you can do with GCP-captured data, the difficulty of the transition really depends on the type and level of accuracy your are needing. If all you need is map to map relative accuracy then the transition is moderate. You can rely solely on the native coordinates that are provided by a GNSS solution. The transition becomes more difficult if you are trying to align to a specific site or surface coordinate system and CAD files that may not be a 1:1 transformation from native GNSS data.

Adam - What brand/model of drones are you all flying?
We have Standard and PPK Phantom 4 Pros as well as standard and PPK Yuneec H520’s.

Grant - What are your thoughts on the current hardware solutions available? What needs to be improved to increase your efficiency?
In my opinion the future is a hexacopter form and there needs to be a middle-ground between the Phantom and the Inspire/Matrice sizes. Something like the Yuneec H520 is a good size. Battery life is always going to be one of the biggest roadblocks, but as camera resolution increase we will be able to gain a little efficiency by flying a little higher. The Phantom 4 Pro has had a practical flight time of 20 minutes with a 5800mah battery whereas the Yuneec has a larger 7900mah battery and even though it and the battery are larger it get 27 minutes of flight time. Another high-end feature of the H520 is a hot-swappable camera system.

Jared - I’ve got a few questions, for importing designs does it have to be a surface or can you import linework from civil 3d?
The import has to be a surface, but a surface can be created from 3D linework. DXF import is coming which should make it much easier to share triangulated surfaces as pretty much any design software can read and directly save\export to a DXF file.

Jared - Do you do anything with the accuracy report from DroneDeploy?
The accuracy report is our primary check on the data ensuring that we have good overlaps, that the GCP’s were tagged correctly and that the checkpoints are within tolerance, but it is also our backup documentation in case the quality of the map ever comes into question so a copy of it is saved in the mission of the project management folder.

Jared - Do you find a difference in the accuracy of stockpiles/cut and fills flying using enhanced 3d vs not?
There is typical not more than a 2-3% variance between the two methods, but if you are using GCP’s it is not recommended to use obliques images because of the extended FOV and distortion and degraded quality of the imagery that result from oblique views.

Kellen - Can you describe your workflow with the Emlid unit?
The workflow will depend on if you are looking for relative map to map accuracy, absolute global GNSS accuracy or site/surface accuracy. With map to map the base station can be setup anywhere and allowed to average in. Then location can commence with the rover. or you can the rover off a CORS or NTRIP service. As long as that final base point is recorded any following surveys will be relative. For global GNSS accuracy the base point needs to be averaged in with corrections from an official source like a CORS or NTRIP services. For site/surface accuracy the base can be setup anywhere and a localization process needs to happen in order to distribute the error due to the difference between planned and actual conditions of the control network.

Jeff - What rules of thumb for most efficient placement of GCPs given site relief, slopes, etc.?
Rule of thumb is that you need at least 4 GCP’s preferably in an arrangement that completely encompasses the site. Then it depends on whether or not you are using a PPK/RTK solution on how many additional points are needed.

Ben - Do you see a benefit to using the drone to analyze cut / fill compared to counting trucks, if so where?
Counting trucks and trusting that everything is loaded and documented correctly has worked for years… or has it? There is allot of room for error and normally they do pretty well, but you can only account for the material that you have tickets for and the action of totaling tickets and then reporting on that is time consuming. Plus allot of hands have to touch the information. Probably a task that takes up time that superintendents could be doing something allot more valuable with. A typical site can be flown in 20 minutes and the full values of every stockpile can be known within a couple of hours and distributed to the whole team. On top of that you get a map and can analyze production between maps. Having the drone takeoffs is also good backup on what the tickets report.

Ken - for RTK equip birds, what’s the very min no of req. GCPs? Will one still get us CM accuracy?
If you don’t need to be exactly align to site coordinates in order to be align with CAD files then one GCP can be used to utilize the single-point calibration tool, but I would recommend having a couple of checkpoints on opposite ends of the site that you can fine tune your adjustment too in order to prorate the error. If you want to absolutely align to design files then you will need to use a minimum of 4 and the standard GCP tagging process. A site become larger you may use as many as 10, but in our calculators that would be up to about a 500 acre tract.

Abhijeet - Hi guys, from your experience do you think conventional topo can be replaced by drone & photogrammetry?
It depends on the subject and the condition of the ground. A standard topographical survey on fairly clear ground can absolutely replace conventional surveying, but as the area becomes more covered with vegetation we have to supplement the drone survey with on the ground shots. If there is so much vegetation that allot of ground shots are required then we usually wait to top it until after it is cleared or shredded. LIDAR is another option, but it is currently very expensive and not cost effective for most companies.

Adam - Does anyone have experience complying with government regulations that do not allow interaction with Chinese servers/software?
Yuneec has an H520 model that is certified and I believe Skydio and Autel are both considered made in the USA, but I do not know their certification status.

Bill - What are the recommendations for control setup on tall material piles and trenches: Do you recommend targets on pile tops and targets in a trench?
I guess it depends on how tall you are talking, but normally unless the areas become larger than an acre or two then GCP’s are not needed for simple volumetric calculations. I think once they get above 100ft or so then some oblique images may be beneficial. You usually want to stay at least 750100ft off the subject in order to maintain that you are getting proper overlaps and I have seen conditions where altitude is limited and that can’t be achieved so obliques were necessary. Ground Control Points are most beneficial when there is a large area to be covered. They align the entire subject in rotation, but also alleviates any doming or bowling that may occur across larger spaces.