摘要
基于一种基于人工神经网络的货物列车超、偏载车号自动识别系统 ,重点讨论了车号区域定位分割方法、智能字符切割技术 ,具有选择注意参数的模板匹配神经网络(SATMNN)及采用多级混合集成分类器字符识别方案。实际应用表明 ,该系统性能良好 ,工作稳定 ,车号区域定位正确率大于 99 8% ,字符识别正确率大于 96 5%。
An automatic overloaded or unbalanced freight wagon license recognition system based on neural networks is presented. The detail discussions about automatically wagon license string lo calization, intelligent character extraction technology, selective attention tem plate matching neural network (SATMNN) for character recognition and hierarchica l hybrid integrated classifier are given in this paper. The system shows the goo d performance and robustness in the practical use, the proper wagon license stri ng localization rate is above 99.8 % and the character recognition accuracy reac hes 96.5%.
出处
《高技术通讯》
EI
CAS
CSCD
2000年第6期44-47,109,共5页
Chinese High Technology Letters
基金
国家自然科学基金!( 69772 0 0 2 )资助项目
关键词
人工神经网络
图像分割
字符识别
货物列车
Artificial neural networks, Image segmentation, String localization, Character recognition
