Given a set of local features e.g. AKAZE is there a principled way to select the strongest set of features? I note that MATLAB now has a selection function Select points with strongest metrics - MATLAB selectStrongest - MathWorks United Kingdom but don’t know how it works/if it is actually useful. Any ideas of a good approach?
you’d have to ask Mathworks to clarify. we can only speculate.
or you could learn about AKAZE and then you can speculate what “strong” means in relation to an AKAZE descriptor. it’s probably the ratio of strongest histogram bin to average of them all, or whatever else describes “good” feature points.
look at goodFeaturesToTrack and its theoretical foundation for more inspiration.