Support Website Contact Support Blog

The implication from the choice of check point


I have a question about the accuracy affected by the choice of check point
In my survey, I chose 5 check point out of total 15 GCP.
actually, when I chose them randomly (half is located in the side, the others in the middle of area), I got around 0.02-0.03 accuracy. After that, I heard that check point could be better when ones in the middle of area are selected because of interpolation show better accuracy than extrapolation.
Therefore, I tried to conduct two other process: one is the selection points only in the middle of area, the other is the selection points in the corner or side of area.
Surprisingly, the first process (0.032-0.035) is better than the second (0.031-0.040), but the original method I did with randomly chosen points shows the best accuracy.
I can’t explain why the results is like this…

could you tell me the reason, or is any principle to choose control point for optimal accuracy?

Control points should be spread around the site ensuring that the outer limits are tied down and also the middle of the site. Check points should then be spread around between the control points. This is a generalization, as will vary a little with the shape and constraints of the site.

For truly independent check points run a project with the control points, and then add the check points taking note of the error associated with the point. This is the true error. This will hopefully reflect the Pix4d report but some bias may be introduced once the checkpoints are added to the processing as they also work like MTPs.

1 Like