License Plate Recognition for Moving Vehicles Using a Moving Camera

This paper is dedicated to a license plate recognition (LPR) system for moving vehicles by using car video camera. The proposed LPR method mainly consists of preprocessing, plate location, and character segmentation & recognition. At irst, the possible regions of license plate are enhanced from the captured images through the proposed edge detection method and gradient-based binarization. Then, the correct plate regions are selected by analyzing the horizontal projection and the corner distribution. A vertical Sobel processing is performed on the segmented license-plate region and then the proposed weighted-binarization method is employed to segment each character of the license, followed by the skew correction. Finally, a probabilistic neural network (PNN) technique is applied to recognize each segmented character. Experimental results show that the accuracy rates of license-plate location and license-plate recognition can achieve 91.7% and 88.5%, respectively.