You could try ismember with the 'rows' option. But this will require them to be equal not similar. Alternatively, you could write something that implements an are rows close type of thing. We could help you with that if you could provide a dataset for us to work with.
More
Here is how I would approach this. Basically we want to take the sqrt sum of squared differences between each row combination. I would do this using the following:
MyDists = reshape(sqrt(sum(bsxfun(@minus,mB1,reshape(mB2.',1,3,[])).^2,2)),size(mB1,1),size(mB2,1));
Now MyDists will contain the distance between each rows in a matrix format.
If you put a threshold on this you can then determine which rows/columns are within a certain distance of each other.
