What does control points look like in viewer ? How spread do they need to be ?

Question

I have a question for you . As I am working on using ERDAS Autosync to orthorectify from an ungeoreferenced aerial photos to known reference image. I have a DEM that to go with it as well.

I understand that in order to use “Tie” points or control points need to have them to be spread out. What I would like to see is an example of control points? I have been in searching for this information on google and I find very limited on it. What I would like to find the information that I prefer to see them in pictures or they can come from the research papers to show what control points looks like. It will help me to understand how spread-out they need to be.

Here is the example I found :

http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.516.2917&rep=rep1&type=pdf

Answer 1

Imagine your image is a flat piece of paper and you are trying to attach it to the surface of a spherical object with as little folds/wrinkles in the paper as possible by pressing push pins through the paper and into the sphere (these are your control points). You could exclusively add a bunch of push pins to the middle of the page, this will do a good job with the middle of the paper but will leave distortions around the border. In actuality, the best practice to stick the piece of the paper to the sphere is to use as many evenly spread out push pins because this will minimize the local distortion through your orthorectified image.

This example speaks to the true challenge of orthorectification, the challenge of fitting a 2D image onto a 3D space. In terms of number of GCPs to use it's really dependent on your use case and personal preference. In school I was told 30+ is good enough, but I've used anywhere from 4-60 depending on the factors related to the use-case such as image size, time requirements, accuracy requirements etc.