hi, i am an opencv newbie with what i am hoping is a simple task for openCV. i am looking to match an unknown "silhouette" (i believe it should be called - a 2-bit black & white image, as below, on the left) against a large (~10,000) existing set of similar images (such as on the right). i am really only interested in matching the _leading edge_ (the bumpy top part), however, the images may be slightly _rotated_ or _skewed_ relative to one another, and have slight variations (such as can be seen on the right side of the example images where there are extra notches etc). a _ranking/confidence_ score would be ideal, and fwiw there would likely be **multiple** (grouped) examples of a given pattern (e.g. as each new match is made, the new image would serve as a new example of *that pattern* for future mapping). so that an unknown image matching well against _several_ patterns in the same group (i.e. known same pattern) would substantially increase the confidence of matching that (general) pattern.
any tips/pointers much appreciated. am hoping there is already a lot of work done in opencv toward this end. thanks!
-jon
↧