Malaysian License Plate Recognition Using Artificial Neural Networks and Evolutionally Computation

In this paper, a license plate recognition method using neural networks and genetic algorithms is proposed. Our system assumes that license plate localization has already been accomplished. We combine two methods into a hybrid system to recognize the license plate’s characters. In the first method, we train neural networks to recognize the characters and then apply a genetic algorithm to guide the neural network in the search. The genetic algorithm is necessary because of size and orientation differences between the neural network training samples and the test characters. The other method is template matching that also uses the genetic algorithm to guide its search for same reasons as in the first method. This work is performed with speed and accuracy in mind, aiming at the online usage of the system. The final system accuracy achieved is 97.3 %.