I look at a number of pointlike objects on a 2D plane and measure the straight line distances between them. For example:
These measurements contain individual errors. How can I measure the overall error in the mesh?
I could do this with physical simulation, by placing vertices at first-estimate coordinates and having the edges be springs, then let it settle iteratively to a minimum-energy configuration. I could then conclude that this configuration is "true" and compare the initial and final edge lengths.
That's one option, but it has a number of issues. I am wondering if, given some chosen error function, there is another, more analytical way to measure the overall error in such a mesh?
(I guess I'm asking if we can conclude whether or not the proposed measurements plausibly fit some true model, to some tolerance.)