电子学报 ›› 2018, Vol. 46 ›› Issue (9): 2188-2193.DOI: 10.3969/j.issn.0372-2112.2018.09.020

• 学术论文 • 上一篇    下一篇

计算机多尺度辅助定位车牌算法

魏亭1, 邱实2, 李晨3, 王锐4   

  1. 1. 西安工业大学, 陕西西安 710021;
    2. 中国科学院西安光学精密机械研究所, 陕西西安 710119;
    3. 中国科学院深海科学与工程研究所, 海南三亚 572000;
    4. 西安市公安局, 陕西西安 710002
  • 收稿日期:2017-05-27 修回日期:2017-09-26 出版日期:2018-09-25 发布日期:2018-09-25
  • 通讯作者: 邱实
  • 作者简介:魏亭 男,1975年生,西安工业大学讲师,硕士,研究方向:智能检测与控制、计算机控制.E-mail:83094738@qq.com;李晨 女,1991年生,中国科学院三亚深海科学与工程研究所助理工程师,硕士,研究方向:图像处理、模式识别.E-mail:lic@idsse.ac.cn;王锐 男,1976年生,西安市公安局警官,硕士,研究方向:计算机、交通管理、信息自动化.
  • 基金资助:
    中国科学院"西部之光"项目(No.XAB2016B23),中国科学院光谱成像重点实验室开放基金项目(No.LSIT201717G)

License Plate Detection Algorithm Based on Computer Multi Scale Assist

WEI Ting1, QIU Shi2, LI Chen3, WANG Rui4   

  1. 1. Xi'an Technological University, Xi'an, Shaanxi 710021, China;
    2. Xi'an Institute of Optics and Precision Mechanics of CAS, Xi'an, Shaanxi 710119, China;
    3. Institute of Deep-sea Science and Engineering of CAS, Sanya, Hainan 572000, China;
    4. Xi'an Municipal Public Security Bureau, Xi'an, Shaanxi 710002, China
  • Received:2017-05-27 Revised:2017-09-26 Online:2018-09-25 Published:2018-09-25

摘要: 违法占道拍摄出的单帧车辆图像具有数据量大、时效性强,检测环境复杂等特点.对其检测需要花费大量的人力与物力.并且人们在定位过程中,无法避免因经验、疲劳等方面的干扰,导致遗漏和错误定位.为此本文从视觉感知角度提出计算机多尺度辅助定位车牌算法.模拟视觉感知原理,从车辆特征、纹理特征、颜色特征尺度,逐次聚焦至车牌所在区域.提出了完整的单帧图像车牌定位流程.并且提出基于边界对的车牌区域准确定位算法.通过对实拍的交通图像实验,表明本算法对于正对的车辆有较高的准确率,符合人类视觉感知的过程可实时的对图像进行车牌检测,可同时检测单幅图片的多个车牌.但对于光线过暗、过强或者颜色失真的情况,仍需要进一步的研究.

关键词: 计算机辅助定位, 视觉感知, 车牌, 多尺度, 边缘对

Abstract: There are the characters of large data,strong effectiveness and complex environment in a single-frame image of illegal encroaching vehicles,the detection of illegal encroaching vehicles requires amount of experimental work and labor.In the process of positioning license plates,people cannot avoid some interference such as experience,fatigue,which results in missing and misplacing license plates.The computer-assisted location algorithm of license plates is proposed from the perspective of visual perception,and used the method of multiple scales in this paper.Simulating the visual perception principle,the computer successively focused to the license plate area according to vehicle characteristics,texture features,and color characteristics scale.We put forward a complete license plate positioning process for the single-frame image and propose an accurate location algorithm for license plate area based on boundary pair.The experiments on real traffic images show that the proposed algorithm has a higher accuracy rate when the image is acquired from the head or rear of vehicles,and is consistent with the process of human visual perception,and can detect multiple license plates on the single image at the same time.But further study need to be carried in the situation of too dark environment,too bright light or color distortion.

Key words: computer assisted location, visual perception, license plate, multi-scale, boundary pair

中图分类号: