Vietnam License Plate Recognition System based on Edge Detection and Neural Networks

This paper proposed an improved Automatic License Plate Recognition (ALPR) system for all types of Vietnam license plates (LP), which consists of three modules: license plate location (LPL), characters segmentation and characters recognition. In the LPL module, we have combined edge detection, image subtraction, mathematical morphology, radon transform, interpolation and specific characteristics of Vietnam LP to locate exact LP region. In the characters segmentation module, we have used peak-to-valley method and statistical parameters of Vietnam LP to segment characters & numbers in one-row & two-row types of Vietnam LP. In the characters recognition module, we have used a Multilayer Perceptron (MLP) neural network and back-propagation (BP) algorithm to recognize characters & numbers of Vietnam LP, two MLP networks used independent for characters & numbers and MLP has trained with noises in the training task. Character & number images processed by the pre-processing task, which obtained high quality of character & number images for the using network task to improve accuracy of the system. We have implemented our ALPR system on 700 Vietnam vehicle images taken from actual system with different conditions such as lightening conditions (night & day), license angles, illumination, size and type, colors and reflected light. Our method is more effective than some existing methods and efficiency of computing time & accuracy is improved and very satisfied for all types of Vietnam LP and Vietnam environment.