摘要
提出了一种新的手写数字识别方法,利用传统的Radon变换,找到了一种新的旋转不变特征,最后采用BP神经网络分类器进行分类。实验表明,该方法不仅具有93.89%的高识别率,而且对字符旋转具有很好的鲁棒性。
A novel rotation-invariant handwritten numeral recognition is presented based on Radon transform and BP neural network.A Radon transform is firstly implemented to a handwritten numeral image to extract the rotation-invariant features and then, the BP network classifier is employed to classify it.Experiments show encouraging results.
出处
《计算机工程与应用》
CSCD
北大核心
2008年第30期13-15,共3页
Computer Engineering and Applications
基金
国家自然科学基金 No.60475042
No.10631080~~
关键词
手写数字识别
特征提取
RADON变换
神经网络
handwritten numerals recognition
feature extraction
Radon transform
neural network
