基于投影法和卷积神经网络的手写汉字图像分割研究  

Application of projection method in the recognition of handwritten Chinese characters

作　　者：张莉[1] 孟范泽 刘思霖 冯锐 王钢[2] 蔡靖[1] Zhang Li;Meng Fanze;Liu Silin;Feng Rui;Wang Gang;Cai Jing(College of Instrumentation&Electrical Engineering,Jilin University,Changchun 130026,China;Beihua University,Jilin 132013,China)

机构地区：[1]吉林大学仪器科学与电气工程学院,吉林长春130026 [2]北华大学,吉林吉林132013

出　　处：《电子技术应用》2021年第11期73-75,80,共4页Application of Electronic Technique

基　　金：吉林大学“大学生创新创业训练计划”创新训练项目(201910183696);吉林省教育厅“十三五”科学技术项目(JJKH20200964KJ);吉林省科技发展计划项目(20190303038SF,20190303043SF)。

摘　　要：为提高手写汉字的识别率,针对手写汉字的有效分割,建立了卷积神经网络手写汉字体识别模型,并对投影法和轮廓检测法的适用性进行了对比分析。实验结果显示,相较于轮廓检测法,投影法更适用于手写汉字识别中对文字图像的处理工作,可以实现对所需文字的有效切分,同时简化手写汉字识别网络的设置并提高识别准确率。In order to improve the recognition rate of handwritten Chinese characters and for the effective segmentation of handwritten Chinese characters,this paper established a handwriting Chinese font recognition model based on convolutional neural network,and compared and analyzed the applicability of projection method and contour detection method.The experimental results show that,compared with the contour detection method,the projection method is more suitable for the processing of characters and images in handwritten Chinese character recognition,which can realize the effective segmentation of required characters,simplify the setting of handwritten Chinese character recognition network and improve the recognition accuracy.

关 键 词：手写汉字识别 神经网络 投影法 轮廓检测法 

分 类 号：TN919.82[电子电信—通信与信息系统]