摘要
针对手写数学公式的识别和计算问题,提出了一种基于卷积神经网络的字符训练方法。利用计算机视觉对数学公式图片进行预处理,采用卷积神经网络进行二维矩阵转换,得到了对应的字符符号,通过后缀表达式计算了识别结果。运用Softmax函数训练了字符模型,统计和分析了几种类型的数学公式识别和计算结果。实验结果证明,通过训练字符能有效提高正确率,该方法可为复杂手写数学公式识别和计算提供参考。
Aiming at the recognition and calculation of handwritten mathematical formulas,a character training method based on convolutional neural network is proposed.Computer vision is used to preprocess pictures of mathematical formulas,and convolutional neural network is used to perform two-dimensional matrix transformation,and the corresponding character symbols are obtained,and the recognition results are calculated by suffix expressions.The character model is trained by Softmax function,and the recognition and calculation results of several types of mathematical formulas are counted and analyzed.The experimental results show that the correct rate can be effectively improved by training characters,and the method can provide a reference for the recognition and calculation of complex handwritten mathematical formulas.
作者
蔡宝
周英敏
顾鸿良
CAI Bao;ZHOU Ying-min;GU Hong-liang(Engineering Training and Innovation Education Center,Shanghai Polytechnic University,Shanghai 201209,China;Baidu Shanghai R&D Center,Shanghai 201210,China)
出处
《计算技术与自动化》
2023年第2期114-118,共5页
Computing Technology and Automation
基金
教育部科技发展中心高校产学研创新基金项目(2018C01059)
上海市教育科学研究一般项目(C20039)
上海第二工业大学青年学术骨干培育项目(EGD17XQD41)。
关键词
数学公式
识别和计算
图片预处理
卷积神经网络
后缀表达式
模型训练
mathematical formula
recognition and calculation
picture preprocessing
convolutional neural network
postfix expression
model training
