I have some digital images/photos of traffic scenes and I want to identify
the closest vehicle's number plate in the image and then read out the number
plate. The number plates are black with white numbers and letters on them.
The images are JPG files. Does anyone have any suggestions about which image
processing functions I should use to go about this task? e.g. would edge
functions be best or should I do some sort of thresholding first etc.
Bronwyn.
License plate detection and recognition appears to be one of the
classic computer vision problems. The implementations I have seen are
typically edge-based. (Note that edge detection almost always
involves some kind of thresholding, so that part of the question
becomes an implicit "yes").
There are several things to consider:
1. How will the car's position in the image be detected?
2. How will the license plate's position in the image be
detected?
3. Once the edge detector has extracted the numbers, will you
perform some pattern recognition to identify the numbers? If so, a
Hopfield neural net may be worth considering.
Hope this helps - reply if you wish to discuss further.
Brett
take care