摘要
随着我国电力行业的快速发展,电网作业与运行环节越来越复杂,因此电力仪表盘的智能化监控与抄表变成了迫切需求。由于变电站场景的复杂性,拍摄的仪表盘图片会出现分辨率低,字体角度倾斜,文字明亮不一等特点,文字识别具有难度。针对于以上问题,首先通过添加并改进文本角度矫正模块且引入坐标注意力机制增强图像空间变换的学习能力;其次利用文本生成算法增强模型的泛化能力;最后设计文字识别模块完成识别任务。最终模型在电力仪表盘数据集上准确率取得了90.2%的优异成绩,为变电站电力仪表智能监控与抄表提供了有效的决策依据。
With the rapid development of China’s power industry, the operation and operation links of power grid are becoming more and more complex. Therefore, the intelligent monitoring and meter reading of power instrument panel has become an urgent demand. Due to the complexity of the substation scene, the instrument panel pictures taken will have the characteristics of low resolution, inclined font angle and different brightness of characters, which makes character recognition difficult. To solve the above problems, firstly, the learning ability of image space transformation is enhanced by adding and improving the text angle correction module and introducing the coordinate attention mechanism;Secondly, text generation algorithm is used to enhance the generalization ability of the model;Finally, the character recognition module is designed to complete the recognition task. The accuracy of the final model in the power instrument panel data set has achieved 90.2%, which provides an effective decision-making basis for the intelligent monitoring and meter reading of power instruments in substations.
作者
金林林
李自愿
王康
马涛
董兰芳
JIN Linlin;LI Ziyuan;WANG Kang;MA Tao;DONG Lanfang(Bozhou Electric Power Supply Company,State Grid Anhui Electric Power Company,Bozhou Anhui 236800,China;School of Computer Science and Technology,University of Science and Technology of China,Hefei 230026,China)
出处
《自动化与仪器仪表》
2023年第3期199-203,212,共6页
Automation & Instrumentation
基金
国网安徽省电力有限公司科技项目-智能电网设备监控机器人研究与应用(5212T02001CM)。
关键词
电力仪表盘
文字识别
端到端
自注意力机制
智能监控
power instrument panel
character recognition
end to end
self attention mechanism
intelligent monitoring
作者简介
金林林(1990-),男,安徽阜阳人,工程师。主要研究方向是电力调度运行、无功电压电力系统及其自动化方向。E-mail:jll90315@163.com。
