机构地区: 华南理工大学土木与交通学院
出 处: 《华南理工大学学报(自然科学版)》 2006年第4期19-23,共5页
摘 要: 车牌定位是车牌识别的基础.文中提出了一种准确快速的车牌定位方法.该方法利用三级阈值,由粗到细,首先剔除大量非角点像素,提高角点的定位速度,用大阈值突出显示车牌区域的角点,再根据车牌纹理的角点特性,利用滑动窗粗选车牌区域,然后对粗选车牌区域用小阈值,对车牌区域进行精确定位.试验证实该方法的车牌定位准确率达到98%,平均耗时约20ms. A method for locating the license plate with high speed and great accuracy is put forward for the license plate recognition. In the locating process, a system of three-level thresholds, ranging from roughness to fineness, is adopted to eliminate a number of non-corner pixels first, which speeds up the locating of the plate corners. Then, a big corner detection threshold is used to foreground the corners in the license plate region, and a slid window is used to approximately detect the license plate. Finally, the detected plate is accurately located using a smaller corner detection threshold. Experimental results show that, by the proposed method, the locating accuracy is rated at 98% and the time cost is reduced to about 20ms.
关 键 词: 车牌定位 车牌识别 角点检测 字符角点 角点分布特征
领 域: [自动化与计算机技术] [自动化与计算机技术]