Originally Posted by
keraba2
The trick to ID'ing plates is to take the frames of the video and "stack" them. The letters will emerge as the data piles up along their shapes. It is probably only something an OpenCV hobbyist would do, but in the case of a hit-and-run, I would think that if you gave a copy of the video to the police, they could pull out the plate fairly easily. Googling seems to produce lots of results.
Yes, I've expected that technology should be available. Of course it works best with still frames. But, a rectangular box should be able to be manipulated to a standard view. GIMP has trapezoidal correction.
It might take some serious manipulation to a moving image on a 360 degree view, taken at an angle, and correct it to square, resize, and do the multiple image processing. But, someone has to have made that already.