汽车牌照的提取方法  

Algorithms to Detect Car Plate Position

作　　者：金玲玲[1] 廖芹[1] 汪刘一[2] 

机构地区：[1]华南理工大学应用数学系,广东广州510640 [2]华南农业大学工程学院,广东广州510642

出　　处：《华南理工大学学报（自然科学版）》2002年第7期95-98,共4页Journal of South China University of Technology(Natural Science Edition)

摘　　要：汽车牌照提取是汽车牌照自动识别的关键一步 .本文中提出了两种提取方法 ,即利用扫描行离差数据、有效谷峰点特征及先验知识来初步定位车牌区域 ;采用彩色分割及多级联合混合集成分类器的车牌提取技术 ,通过多层感知器网络对输入彩色图像进行彩色分割 .将这两种方法联合使用对车牌定位准确率高、鲁棒性好 .车牌定位的准确率达99.3% 。Detecting car plate position is a key step in an automatic car plate recognition system. This paper proposes two methods for detecting car plate position from images: One is to utilize the difference data between scanning lines, property of effective peaking points and some pre_knowledge to basically determine the car plate area. Another method is to use color segmentation and multiplayer perceptron classifier. Multi_layer Percetron Networks are used for color image segmentation. Applying both methods together results in a high accurate position, and robustness. The accurate rate has reached as high as 99.3%, which is a good level for practical system.

关 键 词：汽车牌照 提取方法 车牌定位 有效谷峰点 彩色分割 人工神经网络 混合集成分类器 车牌识别 字符提取 

分 类 号：TP391.43[自动化与计算机技术—计算机应用技术]