基于Attention-CTC的自然场景文本识别算法

doi:10.16180/j.cnki.issn1007-7820.2019.12.007

电子科技 ›› 2019, Vol. 32 ›› Issue (12): 32-36.doi: 10.16180/j.cnki.issn1007-7820.2019.12.007

基于Attention-CTC的自然场景文本识别算法

和文杰,刘敬彪,潘勉,吕帅帅

杭州电子科技大学电子信息学院,浙江杭州 310018

收稿日期:2018-12-07 出版日期:2019-12-15 发布日期:2019-12-24
作者简介:和文杰(1994-),男,硕士研究生。研究方向:机器视觉。|刘敬彪(1964-),男,博士,教授。研究方向:电子系统集成技术。
基金资助:
国家自然科学基金(61871164);国家自然科学基金(61501155)

Natural Scene Text Recognition Algorithm Based on Attention-CTC

HE Wenjie,LIU Jingbiao,PAN Mian,LÜ Shuaishuai

School of Electronic and Information,Hangzhou Dianzi University,Hangzhou 310018,China

Received:2018-12-07 Online:2019-12-15 Published:2019-12-24
Supported by:
National Natural Science Foundation of China(61871164);National Natural Science Foundation of China(61501155)

摘要/Abstract

摘要：

针对自然场景下文本识别所存在的字符分割困难、识别精度依赖字典等问题,文中提出了一种基于注意力机制与连接时间分类损失相结合的文本识别算法。利用卷积神经网络与双向长短时期记忆网络实现对图像的特征编码,再使用Attention-CTC结构实现对特征序列的解码,有效解决Attention解码无约束的问题。该算法避免了对标签进行额外对齐预处理和后续语法处理,在加快训练收敛速度的同时显著提高了文本识别率。实验结果表明,该算法对字体模糊、背景复杂的文本图像都具有很好的鲁棒性。

关键词: 文本识别, 连接时间分类, 卷积神经网络, 循环神经网络, 多尺度特征提取, 注意力机制

Abstract:

In order to solve the problems of the difficulty of character segmentation in text recognition and the recognition accuracy dependent on dictionary in natural scene, a text recognition algorithm with attention mechanism and connection time classification loss was proposed. The convolutional neural network and the bidirectional long-term memory network were used to extra the feature of the image. The Attention-CTC structure was used to decode the feature sequence, which effectively solved the problem of Attention decoding unconstrained. The algorithm avoided additional alignment preprocessing and subsequent syntax processing on the tag, which sped up the training convergence rate and significantly improved the text recognition rate. Experimental results showed that the algorithm was robust to texts with complex fonts and complex backgrounds.

Key words: text recognition, connection time classification, convolution neural network, recurrent neural network, multi-scale feature extraction, attention mechanism

中图分类号:

TP391

和文杰,刘敬彪,潘勉,吕帅帅. 基于Attention-CTC的自然场景文本识别算法[J]. 电子科技, 2019, 32(12): 32-36.

HE Wenjie,LIU Jingbiao,PAN Mian,LÜ Shuaishuai. Natural Scene Text Recognition Algorithm Based on Attention-CTC[J]. Electronic Science and Technology, 2019, 32(12): 32-36.

图/表 5

参考文献 18

[1]	Yao C, Bai X, Shi B , et al. Strokelets:a learned multi-scale representation for scene text recognition [C].Columbus: Computer Vision and Pattern Recognition, 2014.
[2]	Jaderberg M, Vedaldi A, Zisserman A . Deep features for text spotting [C].Cham:Computer Vision, 2014.
[3]	薛皓天, 杨晶东, 谈凯德 . 一种改进的BP神经网络在手写体识别上的应用[J]. 电子科技, 2015,28(5):20-23.
	Xue Haotian, Yang Jingdong, Tan Kaide . Application of an improved BP neural network in handwriting recognition[J]. Electronic Science and Technology, 2015,28(5):20-23.
[4]	熊海朋, 陈洋洋, 陈春玮 . 基于卷积神经网络的场景图像文本定位研究[J]. 电子科技, 2018,31(1):50-53.
	Xiong Haipeng, Chen Xiangxiang, Chen Chunwei . Text location in image based on convolution neural network[J]. Electronic Science and Technology, 2018,31(1):50-53.
[5]	Bahdanau D, Chorowski J, Serdyuk D , et al. End-to-end attention-based large vocabulary speech recognition [C]. Shanghai:The 41^st IEEE International Conference on Acoustics, Speech and Signal Processing , 2016.
[6]	Luong M T, Pham H, Manning C D . Effective approaches to attention-based neural machine translation [C].Lisbon: Empirical Methods in Natural Language Processing, 2015.
[7]	Qu S, Xi Y, Ding S . Visual attention based on long-short term memory model for image caption generation [C]. Melbourne:Control & Decision Conference. 2017.
[8]	Bahdanau D, Cho K, Bengio Y . Neural machine translation by jointly learning to align and translate [C].San Diego: International Conference on Learning Representations, 2015.
[9]	Graves A, Gomez F . Connectionist temporal classification: labelling unsegmented sequence data with recurrent neural networks [C].Hong Kong:International Conference on Machine Learning, 2006.
[10]	尹征, 唐春晖, 张轩雄 . 基于改进型稀疏自动编码器的图像识别[J]. 电子科技, 2016,29(1):124-127.
	Yin Zheng, Tang Chunhui, Zhang Xuanxiong . Image recognition based on improved sparse auto-encoder[J]. Electronic Science and Technology, 2016,29(1):124-127.
[11]	Hochreiter S, Schmidhuber J . Long short-term memory[J]. Neural Computation, 1997,9(8):1735-1780. doi: 10.1162/neco.1997.9.8.1735 pmid: 9377276
[12]	Lucas S M, Panaretos A, Sosa L , et al. ICDAR 2003 robust reading competitions[J]. Proceeding of the Icdar, 2003,7(2-3):105-122.
[13]	Ioffe S, Szegedy C . Batch normalization: accelerating deep network training by reducing internal covariate shift [C].Lille Grand Palais:International Conference on Machine Learning, 2015.
[14]	Szegedy C, Vanhoucke V, Ioffe S , et al. Rethinking the inception architecture for computer vision [C].Las Vegas:Computer Vision and Pattern Recognition, 2016.
[15]	Szegedy C, Ioffe S, Vanhoucke V , et al. Inception-v4, inception-resnet and the impact of residual connections on learning [C].San Francisco:The Thirty-First AAAI Conference on Artificial Intelligence, 2017.
[16]	Kim S, Hori T, Watanabe S . Joint CTC-attention based end-to-end speech recognition using multi-task learning [C].New Orleans:The 42^nd IEEE International Conference on Acoustics,Speech and Signal Processing , 2017.
[17]	Hori T, Watanabe S, Zhang Y , et al. Advances in joint CTC-attention based end-to-end speech recognition with a deep CNN encoder and RNN-LM [C].USA:IEEE International Conference, 2017.
[18]	Xu K, Li D, Cassimatis N , et al. LCANet:end-to-end lipreading with cascaded attention-CTC [C].Xi’an:China Automatic Face & Gesture Recognition, 2018.

模型	ICDAR03	IIIT5K	SVT
CRNN	0.76	0.68	0.70
Attention	0.80	0.74	0.77
Proposed (0.2)	0.86	0.76	0.78
Proposed (0.5)	0.81	0.73	0.76
Proposed (0.8)	0.78	0.70	0.71

基于Attention-CTC的自然场景文本识别算法

Natural Scene Text Recognition Algorithm Based on Attention-CTC

RichHTML

PDF (PC)

赞

可视化

摘要/Abstract

引用本文

使用本文

图/表 5

参考文献 18

相关文章 15

Metrics

本文评价

推荐阅读 10

[1]	李辉,王一丞. 基于CNNCIFG-Attention模型的文本情感分类[J]. 电子科技, 2022, 35(2): 46-51.
[2]	程晓雅,张雷. 基于深度学习的有遮挡人脸识别方法研究[J]. 电子科技, 2022, 35(1): 35-39.
[3]	战荫伟,朱百万,杨卓. 一种车脸识别算法的研究与应用[J]. 电子科技, 2021, 34(8): 1-7.
[4]	廖璐明,张伟. 基于改进VGG16网络的混合批量训练交通标志识别[J]. 电子科技, 2021, 34(8): 8-13.
[5]	马立新,豆晨飞,宋晨灿,杨天笑. 基于特征融合卷积神经网络的绝缘子无损检测[J]. 电子科技, 2021, 34(7): 26-30.
[6]	顾伟,李菲菲,陈虬. 基于多特征融合的行人检测方法[J]. 电子科技, 2021, 34(5): 29-34.
[7]	张仕森,孙宪坤,尹玲,李世玺. 基于神经网络的文本标题生成原型系统设计[J]. 电子科技, 2021, 34(5): 35-41.
[8]	闫书豪,乔美英. 基于一维WConv-BiLSTM的轴承故障诊断算法[J]. 电子科技, 2021, 34(4): 75-82.
[9]	薛永杰,巨志勇. 基于改进AlexNet的鱼类识别算法[J]. 电子科技, 2021, 34(4): 12-17.
[10]	葛靖,刘子龙. 基于CNN和LSTM的睡眠呼吸暂停检测算法[J]. 电子科技, 2021, 34(2): 21-26.
[11]	朱斌,刘子龙. 基于新型初始模块的卷积神经网络图像分类方法[J]. 电子科技, 2021, 34(2): 52-56.
[12]	张灿,陈玮,尹钟. 基于弱监督宫颈细胞图像的语义分割方法[J]. 电子科技, 2021, 34(12): 68-74.
[13]	张莹,刘子龙,万伟. 基于Faster R-CNN的无人机车辆目标检测[J]. 电子科技, 2021, 34(11): 11-20.
[14]	李诚,刘昊,蒋希峰,吴军法,韩文刚,高建国. 基于VGG网络的发电机定转子智能诊断算法[J]. 电子科技, 2021, 34(11): 62-66.
[15]	董波,周燕,王永雄. 基于渐进结构感受野和全局注意力的显著性检测[J]. 电子科技, 2021, 34(1): 23-30.