摘要
给出一种全新有效的快速算法。该方法通过合理的阈值将模板向量转化为 (0 ,1,* ) -向量 ,并充分考虑到代表样本与模板之间相关性的不同因素的不同重要性 ,赋以相应的权系数 ,并用遗传算法来确定阈值和权系数。印刷体邮政编码的实验结果表明 ,该算法在大大缩短识别时间的同时 ,识别率可达98.1% ,而相同实验条件下应用传统模板匹配法时的识别率为 92 .1%。
To reduce the computing complexity of printed character recognition and improve recognition rate as well, a new effective algorithm is proposed. Using two reasonable threshold values, the algorithm transforms real template vectors into (0,1,*)-ones. Meanwhile, adequately considering the different weightiness of the four different factors denoting the pertinence between template and unknown sample, corresponding weight coefficients are allocated on them. Genetic algorithm is used to decide all these threshold values and weight coefficients. The method produces 98.1% of the recognition rate, which is better than 92.1% of conventional template-matching method under the identical experimental condition.
出处
《控制与决策》
EI
CSCD
北大核心
2001年第3期296-298,302,共4页
Control and Decision
基金
国家自然科学基金!项目 (6 0 0 75 0 18)
