As a quality inspection vision application, we have images of printed circuit boards. The PCBs are rectangular in shape and the distance from the camera to the PCB is fixed. The relative location of the PCB in the image can shift or even rotate a bit.
To extract features in the PCB, I would think one needs to establish an "origin" using one corner of the PCB, and use the edges of the PCB as the axis of a new coordinate system. All the features/components of the PCB are fixed in this new coordinate system, which would make looking at a certain area of the PCB easier across images. I think they call this 'origin' the anchor point in image process correct? Please tell me if this is a sensible start to this kind of problem.
If so, what tool/algorithm would you use to establish this new coordinate system.