Agisoft Metashape
Agisoft Metashape => General => Topic started by: jailblazers on August 06, 2014, 03:31:54 PM
-
I've read through several of the forums regarding how errors for ground control points are reported and still don't quite understand how the errors are being calculated.
From what I've read, the total error reported at the bottom of each column is essentially RMSE ("square root from the sum of squares and that's all is divided by the number of GCPs"). Is this correct?
I've put my GCP error values into an excel table and no matter how I try to calculate total error for each column (x,y, and z), or the error values in the 5th column, I can't get the numbers even close to matching.
Would someone please help explain how PS calculates these values?
Thanks in advance.
-
Hello,
root ( (sum v^2)/(n))
should work like this.
where
v = estimated - input data (estimated error)
n= number of GCP
the 3D error (5th column) for each point is the spatial pythagoras.
root (dx^2+dy^2+dz^2) = 3D error
BR
Paul
-
Thanks, Porly
Hello,
root ( (sum v^2)/(n))
should work like this.
where
v = estimated - input data (estimated error)
n= number of GCP
the 3D error (5th column) for each point is the spatial pythagoras.
root (dx^2+dy^2+dz^2) = 3D error
BR
Paul
That definitely helps explain it for me (and hopefully others).
However, calculating 3D error using that method is still giving me a slightly different value (0.355 m) than what PS reports (0.339 m). What does d represent in your 3D error equation, pixel size?
Thanks
-
Hello jailblazers,
For point 1: Error = sqrt(X_error ^ 2 + Y_error ^ 2 + Z_error ^ 2), as it is actually the distance in 3D space between estimated and source location (the latter one is defined by input coordinates)
For total error, as Paul has already said: Total_error = sqrt( sum(point_error ^2) / point_number)
Maybe you can post the exact equation on how you've got 0.355?
-
Thanks, Alexey
If you look at my spreadsheet X error= -0.08187, Y error= 0.062554, Z error= -0.323538, and 3D error reported by PS = 0.339547.
When I calculate, SqRoot[(-0.08187^2)+(0.062554^2)+(-0.323538^2)] = 0.354834983
-
Strange, I got the following:
math.sqrt(0.08187 ** 2 + 0.062554 ** 2 + 0.323538 **2)
0.3395475478633294
Maybe omitting minus signs will help?
-
Me too