I'm using emguCV and tesseract with c# to recognize licence plates number, the picture is catched from an IP camera which is 1280x960 of size and 72 ppi, the result was negative,
the first image :
then i test the same programe with an image from my laptop web camera and the result was positive the plate was reconized and the number was read succsfully (the picture is 640x480 of size).
I need to know if the picture size or resolution can influence the licence late detection !?
There are some great answers on image rectification/morphology, making the best use of FindContours
. You basically need to find that rotation matrix that'll make the plate flat. The guy in the answers does exactly that.
Rectangle detection / tracking using OpenCV
FindContours
and canny edge detection
would be a good try for your needs, since your image can't be HSV-filtered.