论文部分内容阅读
本文提出了一种基于外接同心圆结构提取贯穿特征码的自由手写体数字的神经网络识别方法。该方法是用自由手写体数字的外接同心圆来抽取其贯穿特征码，将获得的模式特征训练改进的BP神经网络分类器，从而达到快速分类的目的。将其应用于邮政编码识别系统，单字的识别率达到97％以上，整信的识别率可达到92％以上，得到了令人满意的结果。
In this paper, a new neural network identification method based on free-handwritten numerals for extracting conformable codes through external concentric circles is proposed. In this method, the free-hand-written number of concentric circles is used to extract its traversal pattern and improve the training pattern features of BP neural network classifier so as to achieve rapid classification. When applied to the postal code recognition system, the recognition rate of single word can reach above 97%, and the recognition rate of whole letter can reach more than 92%, and satisfactory results have been obtained.