摘要
车牌定位作为车牌自动识别中的核心步骤,对系统识别是否精确具有重要影响。现实生活中,车牌定位与识别常常受到诸如天气、背景、车牌磨损、图像倾斜等复杂环境下干扰因素的影响。经过对已有算法进行比较,采用快速的水平差分算子边缘检测算法,利用牌照区域内有边缘相对集中和规则的纹理特征对汽车牌照进行快速分割,设计并实现了一种高效、鲁棒性强、对图片噪声具有较强适应性的车牌定位方法,为下一步快速、精确地识别字符打下了良好基础。大量实验结果证明,该车牌定位方法具有有效性和鲁棒性,定位率可达97%。
License-plate location, as a key step to automatic recognition techniques for license plate, now attracts much attention from people. The images contain rich background noises because they are usually taken outdoors. Based on comparison of the existing algorithms, the fast algorithm for edge detection based on difference operator is used, with fast segmentation of vehicle license plate based on the feature of relative concentration and regular texture features to enhance the original vehicle image and boost up the plate area. Thus the fast, flexible and robust vehicle-license-plate location method is designed and realized. Experimental results indicate that this method is of feasibility, applicability and robustness, with a location rate of 97%.
出处
《通信技术》
2017年第6期1181-1187,共7页
Communications Technology
基金
重庆市教委科学技术研究项目(No.KJ1600937)
重庆理工大学青年科研项目星火支持计划(No.2014XH12)
国家自然科学基金项目(No.61202348)~~
关键词
图像预处理
车牌定位
边缘检测
中值滤波
垂直投影
image preprocessing
vehicle-license-plate location
edge detection
medium filtering
vertical projection
