序列图像的多车牌定位算法  

A Multi-license Plate Location Algorithm for Vehicle Sequential Images

作　　者：丛建亭[1] 邱超[1] 袁凌云[2] 

机构地区：[1]哈尔滨工业大学自动化测试与控制系,黑龙江哈尔滨150001 [2]武汉理工大学资源与环境工程学院,湖北武汉430070

出　　处：《公路交通科技》2012年第6期138-143,共6页Journal of Highway and Transportation Research and Development

摘　　要：针对序列图像的多车牌定位问题,提出了一种综合车牌特征与梯度分析的定位算法。该算法使用灰度均衡化与中值滤波算法对图像预处理来消除噪声,利用测试样本图像提取出若干有效的车牌参数作为定位系统的输入,实现了全自动定位车牌的目的。并引入MGD梯度分析方法辅助边缘提取算法来提取车牌候选区域,利用HSV颜色模型,使用主颜色分析方法分析候选区域,并通过车牌字符纹理分析方法做进一步的判断。使用Hough轮廓检测方法计算非文本区域块的直线斜率。该算法定位效果较好,鲁棒性强,有很好的工程应用推广价值。A new multi-license plate location algorithm combined with license plate characteristics and gray gradient analysis to locate vehicle license plate for sequential images was presented. Image preprocessing technique referring to gray equalization and median filtering algorithms was applied to remove noises. Several efficient license plate related parameters were extracted from test sample image database as the input of location system to fulfill process automation. All license plate candidate regions were extracted by edge extraction algorithm and MGD gradient analysis approach. Only those candidate regions which have achieved the success within principal color analysis based on HSV color model and character texture features analysis were considered for a more thorough inspection. Hough boundary detection was employed to calculate the straight line slopes of non-texture regions, The results show that the proposed algorithm has a better location precision, stronger robustness, and good practical application prospect in the future.

关 键 词：智能运输系统 多车牌定位 MGD梯度分析 HSV颜色模型 Hough轮廓检测 

分 类 号：U491[交通运输工程—交通运输规划与管理] TP391.41[交通运输工程—道路与铁道工程]