摘要
手绘电气草图识别输入有益于电路设计者开拓思路,把握全局,提高设计效率,是现代CAD技术的重要发展内容,各类相关成果不断出现,但总还是存在识别率不高、速度较慢,或者手绘约束太多的等问题.结合构成电气符号草图绘制笔画的形状和RBF神经网络逼近特性好,无局部最小,收敛速度较快且拓扑结构简单等优点,该文研究基于RBF神经网络的在线手绘电气草图分类方法.特征按结构特征和关系特征分类处理与应用,并由此构建两级串联分类系统;第一级分类器使用一个RBF神经网络,输入向量为结构特征;第二级分类器包含三个RBF神经网络,输入内容是关系特征向量.试验表明,该方法识别率高,速度快,基本无约束.
The recognition input of hand-drawn electronic component symbol is beneficial to broaden circuit designers＇ vision,let them grasp the overall situation and raise design efficiency.It is important development matters of contemporary CAD technology.All kind＇s of relate achievements appear out frequently. But there always some problems such as the recognition rate is not high,the recognition speed is slow or have many constrains of hand-drawn and so on.Combining the stroke shape of hand-drawn electronic component symbol and the RBF neural networks advantage of optimal approximation capability,not suffering from local minima,fast constringency speed and simple topology,this paper studied the classification method of hand-drawn electronic component symbol based on RBF neural networks.Classify and apply the feature according to structural feature and relationship feature,from which a two levels in series classification system is constructed.The first level classifier of the series classification system used one RBF neural networks, and selected structural feature for his inputs.The second level classifier of the series classification system used three RBF neural networks,and selected relationship feature for his inputs.The experiment showed that the classification method has high recognition rate,fast recognition speed and few constraints of hand-drawn.
出处
《湘潭大学自然科学学报》
CAS
CSCD
北大核心
2010年第4期102-107,共6页
Natural Science Journal of Xiangtan University
基金
湖南省高校创新平台开放基金资助项目(09K040)
湖南省重点学科建设资助项目
关键词
手绘电气草图
在线识别
RBF神经网络
两级串联分类器
hand-drawn electronic component symbol
on-line recognition
RBF neural networks
classifier
