基于Nanodet-Plus的智能抄表系统研究与设计  

Research and design of intelligent meter reading system based on Nanodet-Plus

作　　者：蒋善旗 康兵 倪嘉豪 邓小珍[1] 丁贵立 许志浩 黄智轩 JIANG Shanqi;KANG Bing;NI Jiahao;DENG Xiaozhen;DING Guili;XU Zhihao;HUANG Zhixuan(Nanchang Institute of Technology,Nanchang 330099,China)

机构地区：[1]南昌工程学院,江西南昌330099

出　　处：《电子设计工程》2023年第22期144-148,共5页Electronic Design Engineering

基　　金：江西省教育厅科技项目(GJJ211943);南昌工程学院校级创新创业专项项目(TZB18D028)。

摘　　要：随着电动汽车数量的不断增加,市场上涌现出了大量电动汽车充电管理公司及平台。对于电动汽车充电管理平台而言,台区变压器的容量信息尤为重要。针对第三方充电管理平台无法直接从国家电网处获取台区变压器容量信息的问题,根据在配电站内安装摄像头以及图像识别的方法,结合配电站实际环境对电能表图片的识别效果进行测试。测试结果表明,系统识别出一张电能表图片的时间在27 ms左右,系统对电能表图片中文字的识别率可达95%,系统对电能表图片中数字的识别率可达98%。With the increasing number of electric vehicles,a large number of electric vehicle charging management companies and platforms have emerged in the market.For electric vehicle charging management platform,the capacity information of transformer in the station area is very important.Aiming at the problem that the third-party charging management platform cannot directly obtain the capacity information of transformer in the station area from the State Grid,the recognition effect of the electricity meter picture is tested by installing cameras and image recognition methods in the distribution station.Combined with the actual environment of the distribution station,the recognition effect of the electricity meter picture is tested.The test results show that the time for the system to recognize an electricity meter picture is about 27 ms,and the recognition rate of the system to the words in the electricity meter picture can reach 95%,and the recognition rate of the system to the numbers in the electricity meter picture can reach 98%.

关 键 词：Nanodet-Plus 变压器容量 树莓派 机器学习 电能表 

分 类 号：TN0[电子电信—物理电子学]