I recently added OpenCV to an SFML project to do video processing (working on a motion detection thing for security camera footage).
From a code point of view the setup was basically (this is for visual studio):
#include "opencv2/opencv.hpp"
#ifdef _DEBUG
#pragma comment(lib, "opencv_world460d.lib")
#else
#pragma comment(lib, "opencv_world460.lib")
#endif
The usual stuff needs to be set up as with any library: includes path, library path and dlls next to the project exe.
Of course OpenCV's images aren't directly compatible with SFML, you'll need to convert them if you need images. So far the only easy way I've found is manually looping over each pixel and converting them.