Skip to Main Content
A new recognition system for the car license plate is designed in view of the low accuracy and the lost time of the identification program in a complex environment, such as low and varying light or strange noisy. The algorithm is composed of five steps: the image preprocessing, the position and segmentation of the license plate, the characters cutting, the feature extraction and the character recognition. It extracts the features based on the multi-scale wavelet. The recognition for characters used the RBF neural network. The new adaptive strategy for raising the speed of recognition is proposed by decreasing the number of some existing hidden layers and the dimension of some features from the wavelet. The experiment results of the actual images show that the average recognition rate can reach more than 92% and the average recognition spend is less 0.11 s than the existing RBF algorithm.