use values(features) in a vector to calculate cosine similarity for opencv

Question

I am recently working on a project where in I have extracted some features regarding an image, and want to find if there are any similarities between two images using those features. Here are the list of features that I have extracted:

1. Aspect ratio (width/height)
2. Normalised area (Cropped roi's area to input image area)
3. Centre of the cropped image
   and other features from a single image

Now, I want to feed these values into a vector, and use that vector to find cosine similarity. In Short, use such vector from two images, and find the similarity between them. I know how cross product of two vectors works, but
I want help in storing these images into vector and usage of the vector. Any suggestions would be deeply appreciated.

Answer 1

oh, not too difficult.

step 1: fill your feature(Mat) with numbers, one after the other:

Mat feature; // you could use a std::vector, too, but cv::Mat has the 
             // handy dot-product used below already built in.
feature.push_back(aspect_ratio);
feature.push_back(area);
feature.push_back(center.x);
feature.push_back(center.y);
feature.push_back(more_stuff);
...

step 2: to compare those features, use the cosine norm :

Mat feature_a, feature_b; // composed like above
double ab = feature_a.dot(feature_b);
double aa = feature_a.dot(feature_a);
double bb = feature_b.dot(feature_b);
return -ab / sqrt(aa*bb);