基于误报分级处理的电表信息识别方法  

A Recognition Method of Electricity Meter’s Information Based on Classification of False Alarm

作　　者：张湛梅 ZHANG Zhanmei(China Mobile Communications Group,Guangzhou Guangdong 510000,China)

机构地区：[1]广东移动公司,广东广州510000

出　　处：《信息与电脑》2022年第18期165-168,共4页Information & Computer

摘　　要：随着深度学习技术的不断发展,利用深度学习技术可以识别各类电表的电表读数和电表编码等信息。随着业务的发展,电表类型也越来越多,复杂的正则化规则已经不能适应当前情况,急需优化。文章基于深度学习的光学文字识别(Optical Character Recognition,OCR)技术和正则化技巧改进电表读数和电表编码等关键信息提取的不足,提出了基于误报分级处理的电表信息识别方法,优化电表图片中指定目标的检测精度、OCR精度以及数据提取逻辑等方面技术,解决了电表读数和电表编码识别精度低的问题。With the continuous development of deep learning technology,it is possible to identify information such as meter readings and meter codes of various meters by using deep learning technology.With the development of business,there are more and more types of electricity meters.Complex regularization rules can no longer adapt to the current situation and need to be optimized.Based on the Optical Character Recognition(OCR)technology of deep learning and regularization techniques,this paper improves the shortcomings of key information extraction such as meter reading and meter coding,proposes a meter information recognition method based on false alarm hierarchical processing,and optimizes the detection accuracy,OCR accuracy and data extraction logic of the specified target in the meter chart.The problem of low recognition accuracy of meter reading and meter code is solved.

关 键 词：图像处理 图像增强 号码识别 数据提取 检测精度 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]