The application of a CICA Neural Network on Farsi license plates recognition

In this paper a new license plates recognition method using a Neural Network, trained by Chaotic Imperialistic Algorithms (CICA), is introduced. In this paper the background of the plate image is omitted, the characters are separated, and then the features of the characters are extracted. The features vector is feed into a multi layered perception neural network trained by CICA. Our dataset include 250 Farsi license plate images for train and 50 images for test in which the test images were noisy. The empirical results of the CICA-NN for license plate recognition are compared with the PSO-NN, GA-NN and MLP neural network. The results show that our method is faster and more accurate than the other methods.