论文部分内容阅读
本文探讨一种新的在工程图扫描图象中自动识别字符、符号的方法，该研究是基于点相关的神经网络识别技术，这种方法考虑了各种退化的字符样本，训练一个字符时使用一种新的学习规则来自动完成训练过程，并从一组字符图象样本集中产生每个字符的理想的特征描述。这种方法使学习的复杂度呈常量，并在工程图字符识别中得到实际的应用。
This paper discusses a new method of automatically recognizing characters and symbols in scanned images of a drawing. The study is based on point-related neural network recognition. This method considers various degenerated character samples and is used when training a character A new learning rule to automate the training process and produce an ideal characterization of each character from a set of character image samples. This method makes the complexity of learning constant, and is actually applied in the drawing character recognition.