基于BP神经网络的多样本手写体字符识别

Handwritten Character Recognition of Multi-sample based on BP Artificial Neural Network

手写体字符识别是人机交互领域的一个重要内容,本文基于BP神经网络实现了任意数量字符模版的多字符手写体字符识别。分为以下几步,第一,首先对目标图像进行识别前预处理。包括灰度图像二值化、图像孤立像素滤波、图像膨胀、腐蚀、按字母最小行分割、按字母最小列分割、图像紧缩、归一化等;第二,用处理好的多个样本进行BP神经网络训练。包括BP网络参数的选择、目标结果构建、输入到结果的映射即用样本库进行神经网络学习机的训练;第三,待测字母的识别。包括对图像预处理、字符提取、归一化和送入已训练好的BP网络进行识别。该系统最终实现了95%以上的手写字符识别正确率,有一定的借鉴意义。

Handwritten Character Recognition is an important element in the field of human-computer interaction. This paper achieved a multi-sample handwritten character recognition based on BP neural network. Divided into the following steps: First, the pre-processing of the target image. Including binarization of gray image, the pixel filtering of isolated image, image dilation and corrosion, character segmentation in minimize row and column, image compression, normalization, etc; Second, Training the BP neural network with the processed character image. Including the selection of BP network parameters, building the results, the input mapping to the output(raining the neural network learning machine using sample database); Third, the recognition test of the unknown handwritten character. Including image preprocessing, character extraction, normalization and typing the unknown character to the BP network that has been trained to recognize. Ultimately, the system achieved more than 95% accuracy of the handwritten character recognition, there is certain significance.