摘要
车牌定位结果中往往存在一些误检,为了减少车牌识别后续模块的计算负担,并提升最终识别结果准确率,提出一种基于字符包络分析和笔画宽度统计的伪车牌排除方法。该方法将车牌定位结果图像二值化和反色判断后,对二值图中字符前景的上下轮廓提取包络,统计上下包络线的间距,设定阈值排除一部分伪车牌;再对二值图像各行前景点进行笔画宽度统计,计算得到多个指标,进一步排除更多伪车牌。实验结果表明,该方法能排除复杂场景中的绝大多数伪车牌。
License plate location may be error detected. A character envelope and calculating stroke length is proposed method of eliminating fake license plate based on analyzing to reduce the computational burden of the license plate recognition module and improve the final accuracy results. In this method, after dealing license plate location results with image binaryzation and invert color judgmen upper and lower envelope, and then analyze t, the license binary image extracted envelope and distance between the the separation distance, so that the threshold is able to be set to exclude a part of fake license plates. Then, the stroke width of foreground points of the license binary image is count, and more fake license plate is excluded after calculating various indicators. Experimental results indicate that this method can eliminate most fake license plates in license plate location results.
作者
杨英仓
Yang Yingcang(Guizhou Road Traffic Accident Identification Engineering Technology Research Center, Guiyang 550005, Guizhou, China Guizhou Police Officer Vocational College, Guiyang 550005, Guizhou, China)
出处
《计算机应用与软件》
2017年第3期222-226,共5页
Computer Applications and Software
基金
贵州省道路交通事故鉴定工程技术研究中心开放基金项目(黔道交鉴合G字[2015]10012号)
关键词
车牌定位
车牌误检
伪车牌
字符包络
笔画宽度
License plate position License error detection Fake license plate Character envelope Stroke width
