Accuracy of the map


I have several years of experience with drones but I am a beginner in construction.

I did a mapping test to get the ground clearance of a small piece of land for a client at a height of 200 feet. I used dronedeply to make the map. I used the Location tool to get the terrain heights. I made a few annotations to mark the heights.

I have the results of:
Orthomosaic Resolution: 0.6 in / px
Accuracy: RMSE 1.8ft - X1.7ft - Y2.4ft - Z0.8ft

I have some questions about the result.

1 - According to the map, the ground would have between 406 feet and 419 feet.
The height of the ground does not represent the height of the flight of the drone. Is this normal?
2 - The client asks, the accuracy of the scan for depth / distance (height)?
3 - The client asks, is this equivalent to a laser pointer eg a 200 m pointer to an accuracy of ± 1 mm or 0.005%

Thank you for enlightening me

  1. The elevation shown in the accuracy report is based off the altitude of the drone mission. Example: in the mission you set the altitude at 100’ the photos will show 100’ as the altitude due to it only knows the elevation where it took off at was 0’. Now you can adjust the the elevation by calibrating the map (under Map Details) of a know elevation.

I use a RTK drone and a base station which allows me to set the base station up and it talks to the drone during the flight to give error corrections and altitude information, this information is then added to the photos. So instead of the the drone taking off at elevation 0’ it would tell the drone your at elevation say 152’ and you fly at altitude of 100’, the photo will show the elevation as 252’ in the exif data.

  1. The overall accuracy depends on the drones camera and how you setup and flew the mission. Did you have enough overlap (front/side) was you altitude correct for the mission, the higher the flight the lower the resolution of accuracy, at 100’ altitude you can see a 1" object on the ground at 300’ you can see a 5" object on the ground. with the lower flight I can see the height better than I can at 300’.

  2. Yes and no, it depends on your setup. If you use Ground Control Panels (GCPs) (with accurate survey data on each panel) then the accuracy will be really good, make sure you use enough GCPs.

If you are flying say a Phantom 4 and it takes 300 photos of a 100 Acre site then your Map won’t be accurate.

I currently fly a 150’ with 65% Side and 80% Front overlaps and I generally will get a 99%+ on map stitching and 15 to 20 Photos per pixel and get 14 to 16 images per GCP (16 is the max in the selection). I usually have 6 GCPs per site (I prefer more) when flying.

When I’m flying 400+ Acres I fly at 305’ with a 68% Side and 80% Front overlaps. Flying higher will see more per photo but because of that it would take less passes to complete the mission so I compensate by adding more Side lap which gives me more routes. This ensures I get all the data I need for the site and maintain the Photo to Pixel Ratio I like to have. Example I just drove 4 hrs to fly a 450 Ac site, it took 6 hrs to fly at 305’ with small overlaps of each mission ( I broke up the site into 5 separate missions) and before I left the site I took all the photos and did a Upload process to DroneDeploy so it would create a Blue Dot Map and that showed me that I did not have any holes in my coverage of the site.

So with all that said accuracy can be there as long as you plan properly and use the proper tools. I flew a 900+ Acres site last month for as part of site improvement survey and they used the Map as a backdrop for all the data the surveyors collected on the ground. It was cool to hear the Map aligned with what the surveyors captured on the ground.

Hello Greg,

I use the Mavic 2 pro for mapping.

My flight was at: 185 feet
The area is: 2.37 acres
The overlap was: frontlap 75% sidelap 65%
Stiching image: 100%
Avg images per pixel: 15.58
Map resolution: 0.59 in / pixel
Map Images: 32

As you suggested, I calibrated the height to 185 feet.

What do you use for GCP?

The customer wants to level the site according to the height I provided him. Is it accurate enough to rely on this data?

Thank you for your time.

You can use anything that is at least 1ft in diameter or 1sqft for GCP’s. Depending on the capabilities of the client you don’t necessarily have to have real-world elevations. We do scenarios all the time for private work where we use an assumed elevation (usually just an orthometric spot grade from Google Earth Pro), get the terrain and remodel as desired. That information can then be staked out or put into GPS Machine Control.

1 Like

Hello Michael,

As you mention, I can use 1sqft panels for gcp? But do I need a surveyor to get the coordinates?

Sorry for my ignorance but I want to try to understand the process.

You don’t need an RPLS/Surveyor to give you coordinates as long as someone has some survey or GPS gear. There’s no need to be on exact state playing coordinates if all they are doing is leveling a piece of land. If they are trying to tie into some construction documents or cad files then you will need to do some work to align with them to make sure that there aren’t any scale factors or shifts in the design information. To take Just a raw piece of land and you weren’t relative to its self drone information is just fine. Depending on how complex the final grading is you don’t even necessarily need ground control points.

How big is the tract?

Can you find out the capabilities of the contractor? It would be nice to know if they are going to take the information directly from you and process it or if they are expecting someone to just stake out the service.

I am already working for this company for follow-ups of their demolition site. I will provide them with the information for the height of the ground.

thanks for taking the time to respond

I drive around a site and look for benchmarks that show elevations. As @MichaelL said it don’t have to be exact elevation unless that is part of the deliverables. I fly sites with no known elevation, the only issue is when the client sees the data they want more out of it and tries to compare it to actual data and they won’t match. But they didn’t want that at first but you can Calibrate the map if you know a elevation of something.

If they client wants exact then they should have some surveying done with a benchmark somewhere. I currently use 1’x1’ targets with a white background and a black +, the other side is the inverse color scheme. You could spend money and get something lick the Propeller Aeropoints, those are targets that calibrate themselves to the network and does real-time gps corrections (trying to keep the description somewhat non technical) these GCPs collect and store the data in a log file that then can be processed after the flight using PPK processing methods. Basically the Aeropoints collects data and 3rd party software reads the photo EXIF data and gets the corrected elevations and adjusts the elevation of the photo.

But all this depends on how accurate does your client want it? The cheapest way is to pick a drone launch site and hve the elevation checked by a surveyor and then Calibrate the mission to that elevation.

If there are any known objects in your first flight that can be located later you can always rerun the map with those as GCP’s. I sometimes set some disposable GCP’s anyway, even if I am not going to locate them right there. If you do get a benchmark that you can apply to those even better.

Also, when I have a map that does not require GCP’s and there will be future flights I will create rough GCP’s in QGIS, rerun the map and then use those for control in future maps. This will allow you to do progress calculations moving forward if needed.

A million ways to skin a cat… If you don’t like cats. :hushed:

1 Like

Hey Guys,

Thanks a lot for your time.

1 Like

Make sure you set up on benchmarks in the correct coordinate system. If you just use an arbitrary base station and rover setup, your elevation will not be even close.

To have a better precision, If I do one point GCP with a surveyor to have the elevation, It is correct or it’s better to have 4-5-6 GCP?

Depends on your need. With one GCP you can calibrate the map and relative accuracy across the site will be about 1-2 meters, but you will not have repeatability between maps. With 4 or more you get rotation and scale so your maps will be more relatable. Your accuracies will get to sub-decimeter (4").