I'm trying to implement a license plate detection algorithm, so far I have narrowed it down to a few interest regions:
My next step would be to classify each interest region and ignore the false regions. I was thinking maybe I can check each region for character. If the region contains some characters, then it is a plate, otherwise, it's a false region. How would I go about checking for characters?
Another approach I can think of is to use PCA to determine if a region contains plate, but I have no idea how to do it in OpenCV.
Text detection is not an easy task at all. It may be harder than the whole plate detection you are building. I can suggest you a simple tricky approach: