matlab - take center of mass or average for matched features -

i have application tracking, have player object following photo shows. need following:

1- detect features each frames , match them next frame, use surf

2- calculate average point feature points have estimated step 1

3- calculate distance between average point estimated @ step 2, between each 2 frames.

then able save location matched features,

surfpoints.location 

but still don't know best way center of mass these points, or take average them?

also how filter miss matched points, see there function estimategeometrictransform , function remove many points matched ones ! there approach that?

so let me sum :

you have 2 keypoint arrays, , matching function gives indices of matches in both lists ("keypoint 7 in original list ~ matching keypoint 12 in second")

so question evaluate global shift these local displacements, taking account outliers ?

in case (fitting model given outliers) should ransac song (and eternally funny ransac song)

although algorithm works great, non-deterministic (as involve trying out models based on random samples , evaluating number of outliers)

---

i'll let reading on ransac's theory (simple statistics), let's see how use ransac in case :

your problem : given list of 2d vectors, find best 2d vector minimizes number of "outliers"

the model fitting step picking vector out of list of vector

outliers vectors go "crazy wrong" in direction or norm

also, ransac explained mathworks