廖美英,周俊煌,张勇军
LIAO Meiying,ZHOU Junhuang,ZHANG Yongjun

• 1:广东科学技术职业学院
• 2:广州市奔流电力科技有限公司
• 3:华南理工大学电力学院

摘要(Abstract)：

针对电力工作票的数字化需求，提出了一种基于种子生长策略与DenseNet（稠密连接网络）的智能识别方法。首先，在文本检测阶段，利用种子生长策略，通过选取初始种子点并逐步扩展候选区域，有效提升了对不规则文本区域的定位精度。然后，在文字识别阶段，结合DenseNet的深层特征提取能力与CTC技术的不定长序列对齐机制，提升了字符序列的识别效果。最后，通过实验证明所提方法在设备代码、电气符号等行业专用字符的识别中表现出更高的准确率，显著降低了误识别率，有效满足了电力工作票数字化处理的需求，具有较强的实用价值和应用前景。
In response to the digitalization needs of electrical work permits, an intelligent recognition method based on the seed growth strategy and dense convolutional network(DenseNet) is proposed. Firstly, during text detection, the seed growth strategy is employed to select an initial seed point and gradually expand the candidate areas, effectively improving the localization accuracy for irregular text areas. Then, during text recognition, the method combines DenseNet's deep feature extraction capabilities with the CTC technique's mechanism for aligning variablelength sequences, enhancing the recognition performance of character sequences. Finally, experiments demonstrate that the proposed method achieves higher accuracy in recognizing industry-specific characters, such as equipment codes and electrical symbols, significantly reducing misrecognition rates and effectively meeting the digital processing needs of electrical work permits. The method shows strong practical value and promising application potential.

关键词(KeyWords)： 种子生长策略;深度神经网络;不定长文本识别;电力工作票
seed growth strategy;deep neural network;variable-length text recognition;electrical work permit

Abstract:

Keywords:

基金项目(Foundation): 国家自然科学基金（62376100）

作者(Author): 廖美英,周俊煌,张勇军
LIAO Meiying,ZHOU Junhuang,ZHANG Yongjun

DOI: 10.19585/j.zjdl.202506009

参考文献(References)：

• [1]丘浩，张炜，林翔宇，等.面向电力作业的工作票分割与作业信息提取方法[J].电力科学与技术学报，2022,37(6):198-205.QIU Hao,ZHANG Wei,LIN Xiangyu,et al.Power operation oriented approach of work ticket segmentation and work information extraction method[J].Journal of Electric Power Science and Technology,2022,37(6):198-205.
• [2]高落落.基于深度学习的光学字符识别关键技术研究[D].武汉：华中科技大学，2022.GAO Luoluo.Research on key technologies of optical characterrecognition based on deep learning[D].Wuhan:Huazhong University of Science and Technology,2022.
• [3]张荣泽，王修晖.改进CRNN网络的直播弹幕文本识别[J].计算机工程与应用，2024,60(15):143-149.ZHANG Rongze,WANG Xiuhui.Live barrage text recognition based on improved CRNN network[J]. Computer Engineering and Applications,2024,60(15):143-149.
• [4] PARK J,LEE E,KIM Y,et al.Multi-lingual optical character recognition system using the reinforcement learning of character segmenter[J].IEEE Access,2020,8:174437-174448.
• [5]褚雪汝，陈中，吴聪颖，等.基于深度学习的电气二次图纸语义识别方法[J].浙江电力，2023,42(8):1-11.CHU Xueru,CHEN Zhong,WU Congying,et al. Small target area extraction and semantic recognition method of electrical secondary drawings based on deep learning[J].Zhejiang Electric Power,2023,42(8):1-11.
• [6]王家凯，黄佩卓，李勇乐，等.电力非结构化大文本特征提取研究[J].浙江电力，2024,43(6):117-124.WANG Jiakai,HUANG Peizhuo,LI Yongle,et al. Research on feature extraction of unstructured large power texts[J].Zhejiang Electric Power,2024,43(6):117-124.
• [7]邬蓉蓉，黎新，宾冬梅.基于XGBoost算法的智能电网信息攻击识别模型[J].电测与仪表，2023,60(1):64-70.WU Rongrong,LI Xin,BIN Dongmei. Network attack identification model of smart grid based on XGBoost algorithm[J]. Electrical Measurement&Instrumentation,2023,60(1):64-70.
• [8]陈典，杨楠，孙骁强，等.基于改进聚类算法的电力系统运行方式识别方法[J].电力电容器与无功补偿，2024,45(2):91-99.CHEN Dian,YANG Nan,SUN Xiaoqiang,et al.Identification method of power system operation mode based on improved clustering algorithm[J].Power Capacitor&Reactive Power Compensation,2024,45(2):91-99.
• [9]石天琛，杨烨，刘明光，等.基于有限信息的电动汽车用户充电行为特征识别[J].电力建设，2024,45(10):69-77.SHI Tianchen,YANG Ye,LIU Mingguang,et al.Identification of charging behavior characteristics of electric vehicle users based on limited information[J].Electric Power Construction,2024,45(10):69-77.
• [10]张中文，吐松江·卡日，张紫薇，等.基于双分支特征融合的电力设备缺陷文本挖掘方法[J].高压电器，2024,60(6):188-196.ZHANG Zhongwen,TUSONGJIANG Kari,ZHANG Ziwei,et al. Text mining method for power equipment defects based on two-branch feature Fusion[J].High Voltage Apparatus,2024,60(6):188-196.
• [11]魏伟，苏津磷，李帆，等.基于改进PSENet与CRNN网络的智能电能表文本识别技术研究[J].电测与仪表，2023,60(12):176-181.WEI Wei,SU Jinlin,LI Fan,et al.Research on scene text recognition technology of smart meter based on improved PSENet and CRNN network[J].Electrical Measurement&Instrumentation,2023,60(12):176-181
• [12]潘成成，卢泉，胡立坤，等.基于颜色分离与形态特征分析的保护压板定位及状态识别研究[J].电测与仪表，2023,60(4):71-77.PAN Chengcheng,LU Quan,HU Likun,et al.Research of platen location and status recognition based on color separation and morphological features analysis[J]. Electrical Measurement&Instrumentation,2023,60(4):71-77.
• [13]孙廷玺，方义治，郑晓东，等.电力电缆局放在线监测神经网络自动识别精度的提升方法[J].高压电器，2024,60(7):210-220.SUN Tingxi,FNAG Yizhi,ZHENG Xiaodong,et al.Improvement method of automatic identification accuracy of on-line monitoring neural network for partial discharge of power cable[J]. High Voltage Apparatus,2024,60(7):210-220.
• [14]胡杰，许刚，齐立忠，等.基于知识图谱的输变电工程辅助评审系统架构及关键技术分析[J].电力建设，2023,44(11):104-112.HU Jie,XU Gang,QI Lizhong,et al.Architecture and key technology analysis of power transmission and transformation engineering auxiliary recheck system based on knowledge graph[J]. Electric Power Construction,2023,44(11):104-112.
• [15]吴彬彬，朱雅魁，葛云龙，等.结合模板匹配和深度神经网络的电能表信息识别[J].电测与仪表，2023,60(9):195-200.WU Binbin,ZHU Yakui,GEYunlong,et al. Information recognition method for electricity meter based on template matching and deep neural network[J].Electrical Measurement&Instrumentation,2023,60(9):195-200.
• [16]赵伟达，陈海文，郭陆阳，等.基于YOLO-E与改进OCRNet图像分割的变电站仪表读数自适应识别方法[J].电力建设，2023,44(11):75-85.ZHAO Weida,CHEN Haiwen,GUO Luyang,et al.Substation meter readings and dial information identification method based on YOLO-E and enhanced OCRNet image segmentation[J]. Electric Power Construction,2023,44(11):75-85.
• [17]邢晓敏，董行，林永清，等.结合角点的边缘检测法实现输电线路杆塔标牌图像定位[J].吉林电力，2020,48(3):32-36.XING Xiaomin,DONG Hang,LIN Yongqing,et al.Edge detection method combining corner points to realize the image location of the tower sign in the transmission line[J].Jilin Electric Power,2020,48(3):32-36.
• [18]俞金瑶，王志伟.基于深度学习的视频文本检测识别算法研究进展[J].数据通信，2024(1):42-46.YU Jinyao,WANG Zhiwei.A survey of video text detection and recognition algorithms based on deep learning[J].Data Communications,2024(1):42-46.
• [19] TIAN Z,HUANG W L,HE T,et al. Detecting text in natural image with connectionist text proposal network[C]//Computer Vision-ECCV 2016. Cham:Springer International Publishing,2016:56-72.
• [20] MA J Q,SHAO W Y,YE H,et al. Arbitrary-oriented scene text detection via rotation proposals[J].IEEE Transactions on Multimedia,2018,20(11):3111-3122.
• [21] SHI B G,BAI X,BELONGIE S.Detecting oriented text in natural images by linking segments[C]//2017 IEEE Conference on Computer Vision and Pattern Recognition(CVPR). July 21-26,2017,Honolulu,HI,USA. IEEE,2017:3482-3490.
• [22]任爽，商继财，杨凯，等.改进YOLOv4的绝缘子图像检测模型[J].广东电力，2023,36(1):94-101.REN Shuang,SHANG Jicai,YANG Kai,et al.Insulator image detection model based on improved YOLOv4[J].Guangdong Electric Power,2023,36(1):94-101.
• [23]袁靖，潘明，朱宁.基于改进型PPYOLOE的电力标志牌检测识别技术研究[J].上海电力大学学报，2023,39(4):399-406.YUAN Jing,PAN Ming,ZHU Ning.Research on detection and recognition of power signboard based on improved PPYOLOE[J].Journal of Shanghai University of Electric Power,2023,39(4):399-406.
• [24] NOVITA,SUYANTO,PRASTI EKO Y. Bonferroni mean fuzzy K-nearest neighbors based handwritten Chinese character recognition[C]//2021 International Conference on Data Science and Its Applications(ICoDSA).October 6-7,2021,Bandung,Indonesia.IEEE,2021:118-123.
• [25] BABU M P,G A.OCR based image text to speech conversion using K-nearest neighbors and comparing with fuzzy K-means clustering algorithm[C]//2023 International Conference on Advances in Computing,Communication and Applied Informatics(ACCAI). May 25-26,2023,Chennai,India.IEEE,2023:1-5.
• [26]丁顶.手机拍照下题目分类算法的研究[D].北京：北京邮电大学，2016.DING Ding.Research on topic classification algorithm under mobile phone photography[D].Beijing:Beijing University of Posts and Telecommunications,2016.
• [27] SONG S B,WAN J Q,YANG Z B,et al.Vision-language pre-training for boosting scene text detectors[C]//2022IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR). June 18-24,2022,New Orleans,LA,USA.IEEE,2022:15660-15670.
• [28] 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.
• [29] SU B L,LU S J.Accurate scene text recognition based on recurrent neural network[C]//Computer Vision-ACCV2014. Cham:Springer International Publishing,2015:35-48.
• [30]钟鸣，陶军，阿敏夫，等.基于注意力机制的二次回路端子文本检测与识别方法[J].电力科学与技术学报，2023,38(3):132-139.ZHONG Ming,TAO Jun,A Minfu,et al. Attention mechanism-based text detection and recognition method for secondary circuit terminals[J]. Journal of Electric Power Science and Technology,2023,38(3):132-139.
• [31] LEE C Y,OSINDERO S. Recursive recurrent nets with attention modeling for OCR in the wild[C]//2016 IEEE Conference on Computer Vision and Pattern Recognition(CVPR).June 27-30,2016,Las Vegas,NV,USA.IEEE,2016:2231-2239.
• [32]戴沁璇，罗晓曙，蒙志明，等.一种基于改进稠密卷积神经网络的表情识别方法[J].现代电子技术，2022,45(9):29-34.DAI Qinxuan,LUO Xiaoshu,MENG Zhiming,et al.Expression recognition method based on improved dense convolutional neural network[J]. Modern Electronics Technique,2022,45(9):29-34.
• [33] SHEPLEY A J,FALZON G,KWAN P,et al. Confluence:a robust non-IoU alternative to non-maxima suppression in object detection[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2023,45(10):11561-11574.
• [34] LI C X,LIU W W,GUO R Y,et al.PP-OCRv3:more attempts for the improvement of ultra lightweight OCR system[EB/OL]. 2022:2206.03001. https：//arxiv. org/abs/2206.03001v2.