views:

157

answers:

1

Im a final year university student who is doing a license plate recognition system as my final year project. I want to know when recognizing the characters what suits the best, is it artificial nurel networks(ANN) or optical character recognition(OCR) using pattern matching? Or is there any easy method I can use? All the answered welcomed.

Thanks a lot

A: 

The easiest method by far is normalized cross correlation. However, I would only recommend this method if there is only one license plate font and no perspective warping. On the other hand, more sophisticated classifiers such as neural networks or support vector machines can take a considerable time to implement.

sargon
Ya will looking to this.Thanks a lot...
License plates are usually rectangular. If you locate them, it is trivial to rectify them again (i.e. compensate the perspective warping). Some simple template matching might be sufficient then already, as the character-set of license plates is rather limited. I guess this heavily depends on how the images are captured...
zerm