摘要
将粗糙集理论与神经网络相结合,针对7×5分辨率的大写英文字母,构建了基于RBF网络的字母识别系统,给出了该识别系统的核心算法与核心结构。该系统利用粗糙集中最小决策算法对识别矩阵进行属性约简,减少了大量的计算和数据库存储量,同时提高了系统识别速度和识别率。通过计算机模拟实验,将该识别系统的识别率与标准BP网络算法及改进BP网络算法相比较,证实了该系统的优越性,在有约1/7的像素点受到随机干扰的情况下,该系统识别率仍可达到88%以上。
A method for capital English letters recognition is proposed with the combination of the rough set theory and neural network algorithms. The recognition system based on RBF net method is developed, and the core algorithms and framework of the system are described. The minimal decision method in rough set is applied to attribute reduction of the letter recognition matrix, and the computation amount is reduced greatly. The numeric experiments are performed to testify the validity of this method. The results show that the average recognition ratio ups to about 88% under some random noises.
出处
《计算机应用与软件》
CSCD
北大核心
2008年第11期210-213,共4页
Computer Applications and Software
关键词
字母识别
粗糙集
RBF网络
属性约简
最小决策算法
Letters recognition Rough set RBF net Attribute reduction Minimal decision method
