I used georeferenced images to create component, after refining registration I disabled gps data from images (Use camera priors for georeferencing to NO). Then I added GCPs and used them, 3 of those were set to ground test. GCPs file was formated like :east,north,alt,and RMS accuracies for each east, nort, alt.
Those accuracies were between 10 to 18 mm.
Exported report shows total values X=0.001, Y = 0.009 and Z = 0.010.
My question is this? How can deviations be under 10 mm id the deviations for each gcp is higher.
Hi @Rango986,
for the GCPs the position accuracy is the range where the point could fit. So the resulted values can be in this range and it could be around 0.
Now I understand that the values can be more precise within the range of individual deviations. Does this mean that the relationship between the GCPs can enhance the accuracy of each other? Also, does the final RMS value represent the actual deviations of the 3D model or map in scale and real world, and I can use this value and deliver to client?
Does this mean that the relationship between the GCPs can enhance the accuracy of each other?
I am not sure if this should be defined like this. GCPs influence the model’s precision, not the each other precision. It also depends, what weight were set for the computation and on how many images the GCPs were placed. To get the better precision idea, you should use just test points.
Also, does the final RMS value represent the actual deviations of the 3D model or map in scale and real world, and I can use this value and deliver to client?
RMS measures the average difference between values predicted by a model and the actual values. It seems like all (GCPs and test points) are used to compute these values. I suppose using only test points should be more real.
Understood. So, test points are used to demonstrate the actual deviations, while GCPs ensure the model’s precision remains within the bounds of these true deviations.
Thank you.
For GCPs it is not always true. The model is trying to be fitted to the GCPs so basically therefore are the errors lower for GCPs. But still, they could be out of the precision value.
I have one more question. Why is there a difference between imported and calculated test points? Shouldn’t test points be excluded from the calculations? See image.
Example test 109 last digits in x,
y=.720 calc y=.727
or
z=8.161 cal z=8.173
That’s it, basically. They are excluded, therefore the values are different to the original position. Or, there are coordinates which should be achieved and which are achieved for the test point.
