摘要
提出一种采用彩色分割及多级混合集成分类器的车牌自动识别方法.该方法由彩色分割、目标定位、字符识别及后处理模块组成.采用多层感知器网络(MLPN)对输入彩色图象进行彩色分割,通过投影法分割出潜在的车牌区域并进一步切割出字符,由多级混合集成分类器给出字符识别的初步识别结果及置信度,经后处理得到最终结果.该方法识别正确率高、鲁棒性好,车牌定位正确率达98.6%,字符识别正确率达到95%以上,具有很好的实用技术指标.
An approach for automatic recognition of a vehicle license using color segmentation and hierarchical hybird integrated classifier is presented. This approach consists of color segmentation, object locating, character recognition and post process modules. A multi layer perceptron networks (MLPN) is employed for stable color segmentation. Projection method is used to locate vehicle license plate based on the prior knowledge of fixed ratio of horizontal and vertical length of a plate and to extract characters in the plate. A hierarchical hybrid integrated classifier is used to recognize these characters, then a license database is used to verify recognition result given by hierarchical hybrid integrated classifier in order to improve the reliability of the recognition results. The experimental result show that the proposed approach is excellent in accuracy and robustness, the proper license plate locating rate is above 98.6% and the character recognition accuracy reaches 95%, and can be put into practical use.
出处
《上海交通大学学报》
EI
CAS
CSCD
北大核心
1998年第10期4-9,共6页
Journal of Shanghai Jiaotong University
基金
国防预研基金
关键词
彩色分割
人工神经网络
汽车牌照
自动识别
color segmentation
artificial neural networks
hierarchical hybrid integrated classifier
vehicle license recognition