摘要
提出了基于非线性势函数的一种几何分类方法,设计了利用电势理论来改造决策面判别函数和利用非线性势函数对手写数字进行分类的算法。以该算法为核心构造的分类器可改善针对不同人写字习惯的分类效果。用VC++实现了该算法,结果表明该算法能减小分类时的错误率,对手写体数字的识别有较好的效果,识别率能达到97%。
A geometry classification algorithm based of nonlinear potential function was put forward. The algorithm was designed to reconstruct discriminant function of decision surface by electric potential theory and to classify handwritten digits by non- linear potential function. Classifier based on this algorithm can improve classifying quality according to writing habits of different people, and the algorithm is realized by VC ＋ ＋. Results indicate that the algorithm can reduce classification er- ror rate and has better recognition effect for handwritten digits. The recognition rate can reach 97%.
出处
《东北林业大学学报》
CAS
CSCD
北大核心
2007年第12期79-81,共3页
Journal of Northeast Forestry University
关键词
模式识别
非线性分类器
势函数
手写数字
Image recognition
Nonlinear classifier
Potential function
Handwritten digits
