Research On Double-line License Plate Recognition Algorithm

The license plate recognition technology is an important part of the intelligent transportation system, is widely used in the field of red light automatically detect and capture, parking charge, electronic police, highway toll. Among them, Double-line license plate recognition is one of the major factors influencing the accuracy of license plate recognition system. Double-line license plate recognition technology is through the method of digital image processing and pattern recognition and using computer software technology to extract identification information of the license plate, which also provides the powerful technical guarantee for the implementation of traffic management and control more automatically, in addition, can also alleviate urban road congestion and play a main role in people’s in our daily life.This paper comprehensively and deeply discusses the algorithm of double-line license plate location and characters segmentation and recognition in double-line license plate recognition system. According to the basic link of double-line license plate recognition-license plate detection technology, this paper proposes a double-line license plate location algorithm based on LAB color space and integrated features. In China, double-line license plate is yellow. So the algorithm transform a vehicle image from RGB color space to LAB color space to extract yellow area which can also remove the interference of other color and improve speed. Then mathematical morphology processing and the inherent prior knowledge of double-line license plate such as area, ratio, pixel distribution, etc. are used to remove the interference of other region and locate the license plate region.Secondly, in the part of double-line license plate character segmentation, tilt angle often influence the character segmentation. Thus, a novel license plate character segmentation method based on category-specific extremely region in gray image is proposed. The method can accurately separate characters without tilt correction.Finally, use principal component analysis to reduce the dimension of the sample character, and transported the training sample set to support vector machine classifier for identify the characters.Experiments indicate that the proposed method makes full use of image information to rapidly identification information of double-row license plate in a complicated environment. Moreover the algorithms are robust to noise and harsh natural environment. And also has very good detection result to a certain extent of dirty license plate.