基于BP网络的手写体数字图片特征提取

BP Network Based on Handwritten Digital Image Feature Extraction

随着计算机网络的发展,越来越多的电子产品会用到手写输入。因此,对这些手写图片的识别变得越来越重要。本文使用的是具有一层隐含层的BP网络模型,提取了图像的7个不变矩(Hu矩)特征、均值、方差以及图像的字符势能,用这些字符特征来对图像进行识别。识别的准确率在80%以上。

With the development of computer network, more and more electronic products use handwritten input. Therefore, the identification of these handwritten images becomes increasingly important. This paper uses the network model with one hidden layer, we extract 7 invariant moments (Hu moments), mean, variance and the characters of the image potential, with these characters to identify the images, and the recognition accuracy is above 80%.