I have an image that is an image of one of a set of objects, and I'd like to use SIFT in Open CV to figure out which object it. I have this code:
Mat pic = imread(...);
vector<KeyPoint> picKP;
Mac picDesc;
Mat obj1 = imread(...);
Mat obj2 = imread(...);
Mat obj3 = imread(...);
vector<KeyPoint> obj1KP;
vector<KeyPoint> obj2KP;
vector<KeyPoint> obj3KP;
Mat obj1Desc;
Mat obj2Desc;
Mat obj3Desc;
SIFT sifter = SIFT();
sifter.detect(pic, picKP);
sifter.detect(obj1, obj1KP);
sifter.detect(obj2, obj2KP);
sifter.detect(obj3, obj3KP);
sifter.compute(pic, picKP, picDesc);
sifter.compute(obj1, obj1KP, obj1Desc);
sifter.compute(obj2, obj2KP, obj1Desc);
sifter.compute(obj3, obj3KP, obj1Desc);
FlannBasedMatcher matcher;
vector<DMatch> matches1;
vector<DMatch> matches2;
vector<DMatch> matches3;
matcher.match(picDesc, obj1Desc, matches1);
matcher.match(picDesc, obj2Desc, matches2);
matcher.match(picDesc, obj3Desc, matches3);
It generates a vector<DMatch> to compare my picture with each potential picture. But I don't know what to do with the 3 vector<DMatch>es to figure out which one is the best match.
Note: I'm doing this mostly to try and learn about SIFT, so even though I know better ways are available for my particular dataset, I still want to try to do it with SIFT.
