罗麟,张非,位一鸣,袁海范
LUO Lin,ZHANG Fei,WEI Yiming,YUAN Haifan

• 1:国网浙江省电力有限公司舟山供电公司

摘要(Abstract)：

为提升电力操作票文字识别的准确度,提出了一种CBTR(基于卷积神经网络的文字识别)方法。首先,基于卷积神经网络学习得到非线性映射函数,提升操作票图像的PSNR(峰值信噪比);然后,基于假想笔画、路径签名与8方向特征构建集成卷积神经网络模型,使用简单平均法计算分类结果,克服手写字体识别难题。最后,选择DLQDF, MCDNN, DeepCNet作为基准方法,使用实际运维检修中操作票图像样本集进行算法验证。结果表明CBTR方法能准确识别操作票图像文字,具有显著的性能优势。
To improve the accuracy of text recognition in power operation tickets, the paper proposes a convolutional neural network-based text recognition method(CBTR). Firstly, CBTR learns a non-linear mapping function based on the convolutional neural network to improve the peak signal-to-noise ratio(PSNR) of the ticket images. Secondly, an integrated convolutional neural network model is constructed based on imaginary strokes, path signatures, and 8-direction features. The classification result is calculated by a simple average method to recognize handwritten characters. Finally, DLQDF, MCDNN and DeepCNet are chosen as baseline methods, and image sample sets of the operation tickets in operation and maintenance are used for algorithm verification. The result shows that CBTR can accurately recognize the images and texts in operation tickets and has performance superiority.

关键词(KeyWords)： 操作票;文字识别;手写字体;卷积神经网络
operation tickets;text recognition;handwritten character;convolutional neural network

Abstract:

Keywords:

基金项目(Foundation): 国网浙江省电力有限公司群创项目（5211ZS19000R)

作者(Author): 罗麟,张非,位一鸣,袁海范
LUO Lin,ZHANG Fei,WEI Yiming,YUAN Haifan

DOI: 10.19585/j.zjdl.202004012

参考文献(References)：

• [1]石葆光.基于深度学习的自然场景文字检测与识别方法研究[D].武汉:华中科技大学,2018.
• [2]HINTON G E.Reducing the dimensionality of data with neural networks[J].Science,2006,313(5786):504-507.
• [3]RANZATO M,POULTNEY C,CHOPRA S,et al.Efficient learning of sparse representations with an energy-based moclel[C]//Advances in Neural Information Processing Systems(NIPS 2006).Cambridge:MIT Press,2017.
• [4]CUN Y L,BOSER B,DENKER J S,et al.Handwritten digit recognition with a back-propagation network[J].Advances in neural information processing systems,1997,2(2):396-404.
• [5]LECUN Y,BOTTOU L,BENGIO Y,et al.Gradient-based learning applied to document recognition[J].Proceedings of the IEEE,1998,86(11):2278-2324.
• [6]HOCHREITER S,SCHMIDHUBER J.Long short-term memory[J].Neural Computation,1997,9(8):1735-1780.
• [7]CIRESAN D C,MEIER U,SCHMIDHUBER J.Transfer learning for Latin and C.hinese characters with deep neural networks[C]//The 2012 International Joint C.onference on Neural Networks(IJCNN).New York:IEEE,2012.
• [8]CIRESAN D C,MEIER U,GAMBARDELLA L M,et al.Convolutional neural network committees for handwritten character classification[C]//2011 International C.onference on Document Analysis and Recognition.New York:IEEE,2011.
• [9]CIREAN D,MEIER U,SCHMIDHUBER J.Multi-column deep neural networks for image classification[J].Eprint Arxiv,2012,157(10):3642-3649.
• [10]LECUN Y,BOTTOU L,BENGIO Y,et al.Gradient-based learning applied to document recognition[J].Proceedings of the IEEE,1998,86(11):2278-2324.
• [11]SIMARD P Y,STEINKRAUS D,PLATT J C.Best practices for convolutional neural networks applied to visual document analysis[C].Icdar.2003,3.
• [12]LEUNG K C,LEUNG C H.Recognition of handwritten Chinese characters by combining regularization,Fisher's discriminant and distorted sample generation[C].2009 10th International Conference on Document Analysis and Recognition.IEEE,2009:1026-1030.
• [13]BASTIEN F,BENGIO Y,BERGERON A,et al.Deep selftaught learning for handwritten character recognition[J/OL].(2010-06-03)[2019-12-24].https://arxiv.org/pdf/1009.3589.pdf.
• [14]OKAMOTO M,NAKAMURA A,YAMAMOTO K.Direction-change features of imaginary strokes for on-line handwriting character recognition[C]//Fourteenth Interna tional C.onference on Pattern Recognition.New York:IEEE,1998.
• [15]GRAHAM B.Spatially-sparse convolutional neural networks[J/OL].(2014-09-23)[2019-12-24].https://arxiv.org/pdf/1409.607 0.pdf.
• [16]BAI Z L,HUO Q.A study on the use of 8-directional features for online handwritten Chinese character recognition[C]//Eighth International Conference on Document Analysis and Recognition.New York:IEEE,2005.
• [17]GRAHAM B.Fractional max-pooling[J/OL].(2015-05-13)[2019-12-24].https://arxiv.org/pdf/1412.6071.pdf.
• [18]YANG W,JIN L,TAO D,et al.DropSample:A new training method to enhance deep convolutional neural networks for large-scale unconstrained handwritten C.hinese character recognition[J].Pattern Recognition,2016(58):190-203.
• [19]WU C,FAN W,HE Y,et al.Handwritten character recog nition by alternately trained relaxation convolutional neural network[C]//2014 14th International Conference on Frontiers in Handwriting Recognition.New York:IEEE,2014.
• [20]佟雨兵,张其善,祁云平.基于PSNR与SSIM联合的图像质量评价模型[J].中国图象图形学报,2006,11(12):1758-1763.
• [21]GLOROT X,BORDES A,BENGIO Y.Deep sparse rectifier neural networks[C]//The fourteenth international conference on artificial intelligence and statistics.Fort Lauderdale:AIST ATS,2011.
• [22]GRAHAM B.Sparse arrays of signatures for online char acter recognition[J/OL].(2013-12-08)[2019-12-24] https://arxiv.org/pdf/1308.0371.pdf.
• [23]周志华.机器学习[M].北京:清华大学出版社,2016.
• [24]金连文,钟卓耀,杨钊,等.深度学习在手写汉字识别中的应用综述[J].自动化学报,2016,42(8):1125-1141.
• [25]LIU C L,YIN F,WANG D H,et al.Online and offline handwritten Chinese character recognition:benchmarking on new databases[J].Pattern Recognition,2013,46(1):155-162.