The application and the problem:
I take some pictures of one apartment building at a time.
But it often happens that Reality Caputre only calculates 9.8 meters for every 10 meters distance measured on location.
The attempt to solve the problem:
I have entered measured values by link points and defined the distance.
No remedy exactly the same values as before -
After a long time of googling I have tried several solutions.
coordinate system changed to WGS68 (exact name may differ)
extremely increased the weighting of the link points so that they outperform the GPS data of the images. from 10 to 1000 (is that enough?)
removed extreme faulty images, (orange line very long and at the wrong point)
The wish:
The possibility, even if the scan is not 100% accurate, to rescale manually before export. That it is very, very accurate afterwards. 20cm is too much.1-5cm are acceptable but also not necessarily wanted.
The question: Is there a way around or a solution to the above problem?
Dear user,
you have been trying this correctly I think, but probably with a missing step. After creating the tie points and defining distance between them (ideally the measure size should be quite big to such a big place to get more accuracy) you need to hit the Update button, which will rescale the object.
Ah all right! Thx for your really fast response!
I’m gonna try that right now.
My workflow is as follows.
I import the drone images and start the map wizard.
After that I check if RC has “scanned” the house correctly, using my measurements of the house and the tool Define Distances.
I have now clicked on update and unfortunately there have been no major improvements and the area is still marked with a yellow exclamation mark.
Also no improvement when I clicked on “Start alignment”.
Well basically all you need to do when scanning such house is get as big measure as you can, or measure a side of the house, set up two control points without big errors so they are accurate, define distance between them, type in the distance what you know the size is like in meters 1 equals one meter. Then hit Update and the scene should rescale. You indeed have a triangle next to the distance which means you have either done the procedure wrong is have typed in something strange, check please. When creating the distance you should also tell RC the starting point and the end point ( control points)
I think I’ve already implemented all your tips in the same way, so I took screenshots of the different areas. Maybe then you can see what I did wrong.
By control points do you mean the type “link point” or one of the other two? “Ground test” or “pass point.”
The problem is I don’t have coordinates from the control points, so I always set these points first with the Define Distance tool and then checked and moved them in the individual images if necessary.
Is it possible to get more and more accurate information about the error than just hovering over it with the mouse?
Thank you very much for the effort! It would be really great if I could get the accuracy from 98-99% to 99,5%+.
Cameras:
EDIT: “Removed building of picture by myself”
Selected Distance:
Name, A,B, activation, Kordinatesystem, Defined Distance, calculated distance. EDIT: “Removed building of picture by myself”
Selected Controllpoint:
Name, Actiavation, Type Linking Point, Weight
Controllpoints and Distance(Einschränkungen) accuracity
Well, there are two types of scaling this. One would be as you have done with the Tie points (and that is what I instructed you to do, so correct.) the second one is to use ground control points measured with an accurate GPS, which is a bit harder and more expensive solution. What I would recommend is to get a large meter and do the measurements on it as it is very accurate if it is for the construction purposes. You now are doing it by a side of a house which you have probably measured, that might be quite inaccurate so I would not expect it to be perfect.
Do you think I should get something to take as an example Exactly 5 meters apart, put it on the floor and scan it and use it to define the distance?
X----5m----X
Yes I have measured the side of the house very accurately, I think you think it is too imprecise to define in RC?
And yes, the GPS measurements are not economically good enough for me to buy something like that. Cost/ benefit is not given.
Do you think I should get something to take as an example Exactly 5 meters apart, put it on the floor and scan it and use it to define the distance?
X----5m----X
Yes basically a big construction ruler on which you will then put the control points directly on the thin lines.
Yes I have measured the side of the house very accurately, I think you think it is too imprecise to define in RC?
Yes indeed it may be off a little and even a slight inaccuracy to real word might mess up the scale a lot.
Although I don’t have an optimal solution yet, I tried it on a new building with a flat and exactly 7 meters long, easy to define surface.
the route is exactly 7 meters, so I set the control points very precisely and then defined 7 meters length. RC says it’s 7.072 meters. After I clicked on update and also after I had it completely recalculated, this distance only reduced to 7.068 meters. So still quite a distance from the defined 7 meters.
At the control points I just removed the comma at the 10,00 something from the weight.
I think this is the same problem as with all other scans, that I can’t adjust the scaling strongly by defining the distance, or because of the GPS data the scaling is not adjusted to the distance I specified.
And therefore I don’t think that a better “measurement tool” will help me. Do you have an idea what this is all about or how I can avoid it?
Could it be the Mavic Mini? and a drone with more pixels than 10MP and a more accurate GPS would help? But it’s a cost factor that I won’t spend for experimentation, only if it can really improve the situation im in.
The hardware matters a lot in terms of precision really. A dslr camera with a very good lens without too much distortion would provide better results for example, even though it is out of your case, it is important to mention. You have a small sensor with a wide angle lens providing lesser accuracy and bigger distortion. Also, again the method of measuring the real world on a ground and then using CPs on place where you think you measured it is not accurate. The pointcloud and the mesh could not be the most accurate in this term, that is why I recommended the ruler which is absolutely precise to the fewest pixels. Anyways your result is not that bad for a lower cost drone with GPS and real world measure.
The measurements I took are accurate to within 10 millimeters. 100% And in this test I could determine the CP with 1-2 pixel accuracy. (or as the middle of 4) And yes absolutely the results are still very very good. And even the worst cases when RC said 10 meters is calculated as 9.7 meters, still very good. Since it was always calculated as too small until now, I hoped to be able to correct this by defining the distance.
And I was and am still confused why RC overwrites the defined distance despite the extremely high weight of the CP. And if I leave out the GPS it works, but loses the orientation which is even more important for me.
But your answer has encouraged me to switch to a larger drone in the next few weeks. Here the economy is in the foreground, so it will not be a 1000€+ Drone. Do you think from the pure data, the Mavic Air 2 will give better results?
The whole thing started as an experiment with the Mavic mini and if I can get it to work faster I might be able to make it a professional drone, but for now I have to test it. And unfortunately I have no possibility to rent a big drone for testing purposes.
I absolutely believe you measured it well in real world, what I am saying is that the shape of the object in 3D space of RC might not be the exact same and therefore taking that into account is not very good. Regarding the drones, I do not have an experience with drones myself and can’t give an exact statement to this, but I know that you should prevent having ultra wide angle lens and you can check the accuracy of the drones gps on their site somewhere. The small drones can’t obviously compare to the bigger ones but yes I understand the money factor. Regarding the distance, well gps will always make some call on the resulting scale and transformation in space.
And in this test I could determine the CP with 1-2 pixel accuracy. (or as the middle of 4) And yes absolutely the results are still very very good. And even the worst cases when RC said 10 meters is calculated as 9.7 meters, still very good. Since it was always calculated as too small until now, I hoped to be able to correct this by defining the distance.