一种基于MEAN-SHIFT特征增强的车牌定位方法  

A License Plate Location Method Based on MEAN-SHIFT Feature Enhancement

作　　者：谢以磊 王军号[1] 颜新云 XIE Yi-lei;WANG Jun-hao;YAN Xin-yun(Anhui University of Science and Technology, Huainan 232001, China;Jinling Institute of Technology, Nanjing 211169, China)

机构地区：[1]安徽理工大学计算机科学与工程学院,安徽淮南232001 [2]金陵科技学院网络与通信工程学院,江苏南京211169

出　　处：《金陵科技学院学报》2020年第1期16-20,共5页Journal of Jinling Institute of Technology

基　　金：江苏省高等学校自然科学研究面上项目(19KJB510030)。

摘　　要：复杂场景下的车牌定位算法由于环境的变化导致车牌识别率低、计算效率不高。在对Faster R-CNN算法分析的基础上,结合MEAN-SHIFT聚类算法的特点,提出了一种基于MEAN-SHIFT特征增强的车牌定位方法。该方法采用并行计算的方式,通过增强目标区域的特征,有效提升了复杂场景下车牌定位的效率和准确度。实验表明,该方法能够在多种复杂场景下快速定位车牌照区域,准确率高,具有较好的鲁棒性。The license plate localization algorithm under complex scenarios has a low license plate recognition rate due to changes in the environment,and the calculation efficiency is not high.Based on the analysis of Faster R-CNN algorithm and the characteristics of MEAN-SHIFT clustering algorithm,this paper proposes a method of vehicle license plate positioning based on matures to effectively improve the efficiency of fast and accurate positioning of license plates in complex scenes.MEAN-SHIFT feature enhancement is a method uses parallel computing to enhance the target area.The fear experiments show that this method can quickly locate the license plate area in a variety of complex scenarios,with high accuracy and good robustness.

关 键 词：复杂场景 FASTER R-CNN MEAN-SHIFT 特征增强 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]