[GRASS-user] georectify - what are the forward and backward error

[Glynn Clements]

Moritz Lennert wrote:

> Can someone explain what is the meaning of the forward and (especially)
> backward errors in georectify/g.transform  ?

Given a set of GCPs, each with source and destination coordinates,
best-fit forward and reverse transformations are computed.

The forward error is the distance between the the forward-transformed
source coordinates and the corresponding destination coordinates,
while the backward error is the distance between the source
coordinates and the reverse-transformed destination coordinates.

> Also, is it normal that the backward error is significantly (by one order
> of magnitude) lower than the forward error ?

I don't know about "normal", but it isn't unlikely. The best-fit
polynomial for one transformation may be a much closer fit than for
its inverse. In particular, the inverse of a polynomial containing
2nd- or 3rd-order terms won't itself be a polynomial.

-- 
Glynn Clements <glynn at gclements.plus.com>