计算机技术与发展2018，Vol.28Issue(6)：129-132,4.DOI:10.3969/j.issn.1673-629X.2018.06.029
基于机器学习的多车牌识别算法应用研究
Research on Application of Multiple License Plate Recognition Algorithm Based on Machine Learning
摘要
Abstract
In order to meet the requirements about high accuracy of multi-license plate recognition for license plate recognition system,we propose a SVM ( support vector machine) based on optimization parameters to locate the license plate. Taking into account the impact of the images taken in hazy weather on the license plate recognition,the first thing is to remove the image haze. Then,the Sobel operator is used for vertical edge detection. According to the morphological processing,the candidate license plate contour is determined and the area of license plate is judged initially by the length and width of the external matrix to the candidate license plate contour. In addition,the col-or of the domestic license plate is single,so the localization rate of the license plate is improved to a great extent by combining the locali-zation of HSV color model. The SVM model is trained to judge the candidate license plate and determine the area which conforms to the license plate. At the same time,after the character segmentation of the license plate,the character recognition is carried out by using the ANN model which is trained by the Chinese characters for the license plate. In addition,the system can train the SVM model in a specific scenario in order to provide a training pattern for different scenarios. It is verified that this system can meet the practical application of multi-license plate recognition and its robustness and accuracy increases nearly 20 percent compared to the general model.关键词
机器学习/车牌识别/支持向量机/多层感知机Key words
machine learning/license plate recognition/support vector machine/multilayer perceptron分类
信息技术与安全科学引用本文复制引用
郑顾平,闫勃勃,李刚..基于机器学习的多车牌识别算法应用研究[J].计算机技术与发展,2018,28(6):129-132,4.基金项目
国家自然科学基金(51407076) （51407076）
河北省自然科学基金(F2014502050) （F2014502050）
河北省高等学校科学研究项目(Z2013007) （Z2013007）
中央高校基本科研业务费专项资金(2015ZD28) （2015ZD28）
