Dynamic Detection of a License Plate Using Neural Network.

This paper proposes a new method for detecting a license plate and recognizing its numerical characters regardless of its image size. With this method, numerical characters on the license plate can be picked out using binarization with a feature that a contrast of the license plate is higher than the other part of the image. The numerical characters are detected and recognized using a neural network that has two functions; one judges whether each pattern is a numerical character or not, and the other recognizes what character it iS. The license plate is detected using positions of the detected characters. To show the effectiveness of this method, experiments were performed. In these experiments using 126 pictures that ware taken under various conditions, each numerical character recognition percent is 94.0% and the license plate detection percent 97.6%.