基于字符特征约束的自适应车牌校正提取  

Adaptive license plate correction and extraction based on character feature constraint

作　　者：费继友[1] 谢金路 李花[1] 王英邗 陈东东[2] 

机构地区：[1]大连交通大学动车运用与维护工程学院,大连116028 [2]大连交通大学机械工程学院,大连116028

出　　处：《仪器仪表学报》2016年第3期632-639,共8页Chinese Journal of Scientific Instrument

基　　金：国家自然科学基金(51376028);图家"863"高技术研究发展计划(2013AA041108)项目资助

摘　　要：针对复杂多变环境中难以快速、精确提取车牌的问题,提出了一种利用字符特征智能校正提取车牌的方法。首先通过Gamma校正和Canny算子结合的方法在灰度图中实现自适应阈值边缘检测,解决了分割阈值选取的难题;然后应用字符特征约束条件提取特征轮廓,根据特征轮廓分布规律提取车牌候选区,避免复杂运算的同时提高了定位准确性;最后对候选区进行线性畸变校正并利用行灰度跳变统计实现了车牌真实性验证和精确提取,为后续的识别工作提供了良好条件。对不同环境中随机采集的700幅高清图像进行测试,综合提取准确率为96%,提取车牌字符规整、无多余残留信息。实验结果显示,该方法对图像中车牌状态、背景环境、光照条件等限制极少,具有更广的适用范围和更强的鲁棒性。Aiming at the problem that it is difficult to extract regular license plates rapidly and accurately in complex environment,an intelligent license plate correction and extraction method based on license plate character feature is proposed. Firstly,this method implements adaptive threshold edge detection in the grey scale image by combining Gamma correction and Canny operator to solve the difficulty of segmentation threshold selection. Then,the feature contours are extracted using the license plate character feature constraints,and the candidate area of the license plate is extracted by the distribution rule of the feature contours,which avoids complicated calculation and improves the positioning accuracy. Finally,the linear distortion correction of the candidate area is conducted,and the authenticity verification and accurate extraction of the license plate are achieved using the line gray level jump statistics,which provides a good precondition for the following identification task. The test experiment was carried out on a dataset containing 700 high-resolution images randomly acquired from various environment; the comprehensive extraction rate is 96%; and in the extracted license plates,the characters are clear and neat without redundant information. Experiment results show that the proposed method has little limitation on its usage condition like license plate status,background environment,illumination condition and etc. in the images,and has broad application range and strong robustness.

关 键 词：车牌提取 字符特征 边缘检测 车牌校正 图像增强 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术] TH7[自动化与计算机技术—计算机科学与技术]