基于图像边缘轮廓的银行支票小写金额识别

Courtesy Amount Image Recognition in Bank Check Based on the Edge of Image

银行支票的小写金额识别已有比较成熟的发展,针对模版匹配手写数字识别有太多的不确定性,BP网络手写数字识别训练复杂,参数不易设置等。本文提出了一种新的以手写数字0～9图像四个方向不同边缘轮廓为特征矢量,当前输入图像的特征矢量与标准特征矢量的最小欧几里德距离为判别公式,识别出支票上的小写金额,识别率98.5%。

The courtesy amount recognition in bank check has been developed maturely.For the handwritten digits recognition,the model matching have too many uncertainties,and the BP network is complex for training and difficult to set parameters and so on.In this paper,a new method is proposed to extract the edge of 0～9 handwritten digits image as feature vector, the minimum Euclidean distance betweens the feature vector of input image and the standard feature vector as the discriminating formula,then identify courtesy amount in back check.