本文基于OCR和图像检测技术设计并实现了一个解决盖章文书图像审核耗时、低效、准确率无保障问题的自动审核方法。具体包括三个部分：文字识别、印章识别和表格内容审核。其中文字识别部分包括带有角度的文本检测算法SegLink以及卷积递归神经网络（convolutional recurrent neural network,CRNN）；印章识别部分包括印章识别与提取算法YOLOv3和印章内容识别方法——极坐标变换法；表格内容审核部分根据预设的规则对表格内容进行完备性和正确性检测。实验结果表明，该方法对此类盖章文书图像具有较高的审核准确率。
In this paper, we design and implement an auto-checking method based on OCR and image detection to replace the time-consuming and error-prone manual work. The method consists of three parts: text recognition, seal recognition, and content checking. For text recognition, we utilize the SegLink algorithm for angled text detection and the CRNN algorithm for variable length end-to-end text recognition. For seal recognition, we employ the YOLOv3 algorithm for seal recognition and extraction, along with the polar coordinate transformation method for seal content recognition. The content checking is based on the preset rules to check the completeness and correctness of the content extracted from the form. Experimental result shows that the proposed method achieves high accuracy in checking stamped document image with seals.
[1] 骆蓉, 黄俊, 黎茂锋, 等. 基于Word模板的复杂文档快速生成方法[J]. 计算机应用与软件, 2020, 37(10):57-63. Luo R, Huang J, Li M F, et al. A fast generation method of complex documents based on word template[J]. Computer Applications and Software, 2020, 37(10):57-63. (in Chinese)
[2] Stevens M E. Introduction to the special issue on optical character recognition (OCR)[J]. Pattern Recognition, 1970, 2(3):147-150.
[3] Tian Z, Huang W L, He T, et al. Detecting text in natural image with connectionist text proposal network[C]//European Conference on Computer Vision. Cham:Springer, 2016:56-72.
[4] Zhou X Y, Yao C, Wen H, et al. EAST:an efficient and accurate scene text detector[C]//2017 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2017:2642-2651.
[5] Yin W P, Schütze H, Xiang B, et al. ABCNN:attention-based convolutional neural network for modeling sentence pairs[J]. Transactions of the Association for Computational Linguistics, 2016, 4:259-272.
[6] Shi B G, Bai X, Yao C. An end-to-end trainable neural network for image-based sequence recognition and its application to scene text recognition[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(11):2298-2304.
[7] 欧阳欢, 范大昭, 李东子. 多特征融合决策的发票印章识别[J]. 计算机工程与设计, 2018, 39(9):2842-2847. Ouyang H, Fan D Z, Li D Z. Invoice seal identification based on multi-feature fusion decision[J]. Computer Engineering and Design, 2018, 39(9):2842-2847. (in Chinese)
[8] Shi B G, Bai X, Belongie S. Detecting oriented text in natural images by linking segments[C]//2017 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2017:3482-3490.
[9] Redmon J, Farhadi A. YOLOv3:an incremental improvement[DB/OL]. 2018[2021-12-01]. https://arxiv.org/abs/1804.02767.
[10] Shi X, Chen Z, Wang H, et al. Convolutional LSTM network:a machine learning approach for precipitation nowcasting[DB/OL]. 2015[2021-12-01]. https://arxiv.org/abs/1506.04214.
[11] Wang Z, Li X, Zhou J. Small-footprint keyword spotting using deep neural network and connectionist temporal classifier[DB/OL]. 2017[2021-12-01]. https://arxiv.org/abs/1709.03665.
[12] Everingham M, Van Gool L, Williams C K I, et al. The pascal visual object classes (VOC) challenge[J]. International Journal of Computer Vision, 2010, 88(2):303-338.
