Avoiding GCPs.

[PeteM]

Some premises.

In my work I'm not interested in the georeferencing of the topographic survey and I'm not interested in high precisions.

I need just relative heights and relative distances, even with 10-20 cm error.

Always I got some kind of previous surveys in which large portions of land, unchanged, are measured with great precision.

My idea is to avoid the GCPs and then check or adjust the point cloud based on the pre-existing topographic survey.

Possible ?

 

[Bigriver]

With good GPS signals and no GCP's I get around 1:100 accuracy horizontally. Heights of objects are not as accurate. If that's good enough for your work, go for it. You could also consider using scale constraint points (Pix4D) which do not require survey equipment, but can significantly improve accuracy of your mapping.

 

[PeteM]

You could also consider using scale constraint points (Pix4D)

It would be enough just to lay down on the ground a rigid bar of known length to have a sort of graphic scale directly on the photos.

However, the problem of heights remains.

Knowing the heights of some points could not operate with the same logic to force the scaling of the entire cloud of points under a few centimeters error ?

 

[Bigriver]

Since Scale Constraints do not have an elevation component, they will only help with corrections in the XY direction. However, not knowing exactly how the software adjusts the model, there may be a small amount of correction in the Z direction as well.
If I am concerned with heights of object, I usually try to set GCPs near the lowest and highest elevations. If this is impractical, I have set GCP's on top of a tall survey rod, to at least get some vertical component to the adjustments.

If you are determined not to use precise GCP's, maybe you could figure out a way to set some "approximate" GCPs. As long as the coordinates of the "approximate" GCP's meet or exceed the accuracy you desire (10-20 cm), your adjusted model should come out close to that quality, or at least better than it would with no adjustments. Test it out, and set some check points in the same manner to verify your results.

 

[PeteM]

With Scale Constraints I can adjust the XY error.

Regarding the error Z, my idea would be to take advantage of previous surveys (with many points Z corrected and unchanged) to adjust the current survey (in the zones that has changed compared to the past).

Possible ?

 

[Bigriver]

Possibly. I don't fully understand your statement concerning Z errors. But test your idea against check points to prove your accuracy.

 

[PeteM]

The logic is the same as GCPs.

The difference is that I wanna use a bunch of known coordinate points obtained from previous topographical surveys of the same area.

My perplexity about this method is the fact that while for a GCP I can place the point very precisely on the photo (in the middle of the X) for an unmarked point there is an error of a few centimeters in the placement of the GCP on the photo.

Anyway...

From some experiments I have noticed that, without using any kind of constraint, if the drone proceeds at a very low speed and stops at every point where the photo is taken the error is very low and acceptable for my needs. Unfortunately, flying in this way consumes a lot of battery and after a few minutes you have to land