Ask Your Question

Revision history [back]

click to hide/show revision 1
initial version

You could use matchTemplate or a similar method after masking both images to only the actual knobs.

You can decide which type of knob it is by comparing histograms. Then separate the ROI around the knob, mask it to only the circle, scale the templates to the correct size (not necessary if you're using a scale-invariant matching) while also masking only the actual knobs and try to match them.

Another thing that would probably work pretty well is using a cascade classifier, trained to the template images masked to only the knobs.