- Joined
- Oct 11, 2017
- Messages
- 66
- Reaction score
- 19
- Age
- 51
You can use the UGCS application with which your h520 will do the mission taking into account the relative altitude of the terrain.
I think it needs to be pointed out that you should take these claims of cm accuracies with a grain of salt. While you may get cm relative accuracy to your known point using RTK, your map can only be ~half as accurate as your known point. I've pointed this out before. You may expect accuracies of a quarter of the vertical accuracy of your known point. These are the standards set out by the ASPRS.
If you start doing mapping and photogrammetry work you had best know these things and not oversell, especially if one of your clients ends up taking you to court. It should also be noted, in the States at least, that you may be running afoul of certain state laws for surveying, mapping, and photogrammetry if you are performing these services without a license. Something to think about.
If you start doing mapping and photogrammetry work you had best know these things and not oversell, especially if one of your clients ends up taking you to court. It should also be noted, in the States at least, that you may be running afoul of certain state laws for surveying, mapping, and photogrammetry if you are performing these services without a license. Something to think about.
Indeed, and there is an important distinction between resolution (number of pixels per cm/inch of ground) and accuracy (where the stitched image appears to place each physical feature). Mapping is a mathematically complex process with a lot of parameters - but the quick online apps make it appear that anyone can just fly a drone and produce an 'accurate map'.
I'm taking graduate classes for these things and as Tuna says, it's not just as easy as flying an automated flight and throwing some pictures into DroneDeploy. I'm not discouraging anyone from pursuing aerial mapping but there is a lot to know and certain standards to follow and in some places laws to comply with. Every time you make a qualification of accuracy and precision you open yourself up to a potential lawsuit and monetary damages. I'm not sure that the vast majority of commercial drone pilots out there have even a basic understanding of aerial mapping and photogrammetry. It's a little worrying.
arruntus
Premium Pilot
- Joined
- Aug 19, 2017
- Messages
- 1,851
- Reaction score
- 730
In an area with very high elevation differences you lose resolution, that's what we're talking about. If there is a 30m difference in height, it is noticeable. No program can do anything about it. I think they're talking about different things. A flight plan with ground tracking solves it. A clear example could be a quarry.
Note: The translator is playing me
Last edited:
I appreciate what you are saying, but I am investigating for a company I work for initially to understand what we may be able to achieve for drone surveying and inspection. I will hopefully be trialing aerial mapping for them soon and need to find out as much info as I can. The imagery and topo surveys will be handed to the CAD guys to decide if they are accurate enough for them to use and produce drawings suitable for setting out of certain things, we are not talking millimeters here. I will not be exposing myself to any liabilities.
I never considered I would simply be "throwing some pictures into DroneDeploy"
I never considered I would simply be "throwing some pictures into DroneDeploy"
Understood, and that comment is not necessarily directed towards you... I was speaking more generally about that I see in the field. People try to sell the service without knowing the first thing about it and I feel like that is bad for our industry as a whole.
True, but then some of the blame should lie with those selling the products and "how to start a drone business" videos. It is the same with all new tech as far as I can see.
Thanks for the tip. I guess this app can replace DP which is in current version 1.3 missing a lot of basic things. I do not know yet if you can control camera angle through UGCS but the fact that it can take into account different elevations of a surveyed area is already a great step forward. And I can see they support Yuneec h520.
I have a home built quad that carries a Sony a6300 APSC camera that will follow terrain in missions to maintain a given AGL. But uses a Pixhawk 2.1 flight controller running Arducopter.
You should point the CAD guys to the technical information provided by the stitching software you use. Unless they know how accuracy (and resolution) can vary across a survey, they're likely to measure the wrong thing. It's entirely possible for parts of a survey to be completely accurate whilst others are skewed, warped or mis-scaled. You can even have situations where large 'corner to corner' measurements are accurate across the whole map, but local points 'inside' the map are incorrect. Each survey can show different results.
I don't really want to be banging on about this, but it's a pipeline where each stage needs to have some understanding of how their actions might affect the output. If you're setting out based on a survey, ground control points may be required and so on...
Would you care to expand on the build? A home build is on my long list of projects to consider, currently thinking of a Tart 960 with Pixhawk 2 cube
Sounds like you are already off to a good start. The 960 is a great kit, but will depend on what camera you will carry which will dictate the gimbal. Mine build is very large and spins 21x6" props and 2 6s batteries in series for 48v power. 80a ESCs & 135kv motors. I can fly the Sony a6300 for about 20min.
What wheelbase? I do find the idea slightly daunting, not so much the design and build but the setting up and programming and wonder if I could cope. I was thinking of building a F550 from a kit but I would have no use for it once finished or if I should just jump in at the deep end, I wish there was a bit more reading material on the subject
Gents,
This subject is full of so much misinformation.
The stitching software we use (Pix4D agisoft etc) are matched by points across many photographs before a GCP is encountered - you have no way of confirming how accurate this process is.
Your photogrammetry software may be stating 15-20mm accuracy on the GCP but the GCP then selves are picked out and have a pixel distortion depending on the GSD- they also are usually GNSS measured so there is another 20-30mm.
So the BEST you will be getting is 50mm!! Then it will distort between so depending on the light and shadows, Overlap GSD etc you can not guarantee a map precision!!
Unless you have GCP in every image!!!
I have carried out many experiments comparing to laser scan / total station and GPS surveys.. depending on GCP placement the level accuracy varies immensely - I have seen level errors of up to 3 meters!!! The best I have seen is +-75mm and some areas towards the edges were double this.
A huge amount of experience an knowledge is required in geospatial methods and techniques before you can begin to be confident with the tolerances.
I do not feel that this technology is to be used for Topographical survey - it is simply not accurate enough. But as an additional deliverable (2D ortho overlay) or for volumetric survey it has advantages..
This subject is full of so much misinformation.
The stitching software we use (Pix4D agisoft etc) are matched by points across many photographs before a GCP is encountered - you have no way of confirming how accurate this process is.
Your photogrammetry software may be stating 15-20mm accuracy on the GCP but the GCP then selves are picked out and have a pixel distortion depending on the GSD- they also are usually GNSS measured so there is another 20-30mm.
So the BEST you will be getting is 50mm!! Then it will distort between so depending on the light and shadows, Overlap GSD etc you can not guarantee a map precision!!
Unless you have GCP in every image!!!
I have carried out many experiments comparing to laser scan / total station and GPS surveys.. depending on GCP placement the level accuracy varies immensely - I have seen level errors of up to 3 meters!!! The best I have seen is +-75mm and some areas towards the edges were double this.
A huge amount of experience an knowledge is required in geospatial methods and techniques before you can begin to be confident with the tolerances.
I do not feel that this technology is to be used for Topographical survey - it is simply not accurate enough. But as an additional deliverable (2D ortho overlay) or for volumetric survey it has advantages..
There are many build threads out there for PH builds. Setting up the bird to fly has also become easier with more automated tools in arducopter, so I wouldn't be too concerned about that. The main thing to do first is determine payload and All Up Weight (AUW). Then obtain the necessary hardware to match that weight.
arruntus
Premium Pilot
- Joined
- Aug 19, 2017
- Messages
- 1,851
- Reaction score
- 730
I don't agree with you on this. When the images are put together, hundreds of thousands of points are compared, which makes them perfectly linked (keypoints). Another thing is the model that is not correctly georeferenced. It may be tilted, not having the correct scale, etc. This is where the GPC's or RTK georeferencing comes in.
Since the model is geometrically correct, matching it to reality depends on the accuracy of the reference points provided to the photogrammetry program. I have seen comparisons between photogrammetric models, models with total station and using GPC's and the variation between the control points of the different techniques only in very few cases reached 10mm. I have found several studies that show what I am saying.
For example:
How does photogrammetry measure up? A comparison - Pix4D
Do RTK/PPK drones give you better results than using GCPs? - Pix4D
These two videos are in Spanish, they are worth watching:
Arruntus these are all tests done by the software vendors.
I am a land surveyor with 20 years experience in mapping. Using GPS and Total Stations.
The amount of work required to get a survey to sub 10mm you would not believe!!!
I have carried out many tests myself and I did encounter a model which had 3m error in height on the top of some areas.
The ground had very significant level differences and only 6 GCP placed. The only way to ensure a good model is to place many more GCP.
I now place approximately 20 on this site. And they have to be placed in certain places to get an accurate model. On a flat smooth site you will get better results.
You will never get 10mm with photogrammetry or even Lidar.. GPS will only get this after Static OBS and post processing for at least 30minutes on a tripod..
When you have completed 25-30 surveys on the same site using GPS and Total Station data to compare with your photogrammetry then you will know.
I survey the same 3 sites every 30 days - and have been doing so for nearly 2 years. The data is never any closer than 50mm in accuracy on the areas which do not change - no matter what is done. Even using the same GCP points you will get a different result.
Similar threads
