Automatic number plate recognition system

Abstract

This thesis implements an automatic license plate recognition system on Vietnamese vehicles which consist of three distinct stages: license plate localization in which we use Hough transform to detect borders of license plates, license plate segmentation into individual characters which comprising of region growing and adaptive thresholding and character recognition by feed forward neural networks. The system works well for images in the range of 5 to 10 meters. We test totally 70 images in which the successful rate for the previous three stages are 71%, 96%, 96% respectively. The system is implemented using Matlab 7.4 on an Intel Core 2 Duo (2.53 GHz) computer. The average processing time is 30 seconds.