Abstract: In order to improve the input efficiency of the experiment report grades in experiment teaching, this paper designs a handwritten character recognition system based on image processing technology, which is able to identify and save student id with VISUAL C++ 6.0 being the compiler environment, MFC as graphics interface development platform, CCD camera for image acquisition, and preprocessing of image based on image recognition theory. This paper also tests both methods of BP neural network and template matching to identify report grades and student’s id and compares the pros and cons of their recognition accuracy and speed. The results show that template matching methods have a lower accuracy rate than BP neural network methods, but are almost 10 times faster, and both of them improve the data input speed greatly compared with the traditional method of manual entry. This system can be applied to quick entry of various types of experimental report grades.

Key words: image processing, template matching, Back Propagation（BP） neural network, handwritten character recognition

摘要： 为提高实验教学中实验报告成绩的录入效率，设计了一个基于图像处理技术的手写字符识别系统。以VISUAL C++ 6.0为编译环境，MFC为图形界面开发平台，通过CCD摄像头进行图像采集，根据图像识别原理对图片进行预处理，并分别采用BP神经网络和模板匹配两种不同方法对实验报告成绩及学号字符进行识别，比较了两种方法在识别准确率和速度方面的优劣。测试结果表明，BP神经网络法比模板匹配法识别的准确率更高，而后者识别速度较后者快10倍左右，自动打分系统较传统的手工录入法大幅度提高了数据输入速度。该系统可以应用于各类实验报告成绩的快速录入。

关键词: 图像处理, 模板匹配, 反向传播（BP）神经网络, 手写字符识别