继保压板状态图像识别嵌入式方案设计与应用  

Relay Platen Status Image Recognition Embedded Solution Design and Application

作　　者：李新海 范德和 孟晨旭 曾令诚 袁拓来 LI Xinhai;FAN Dehe;MENG Chenxu;ZENG Lingcheng;YUAN Tuolai(Zhongshan Power Supply Bureau of Guangdong Power Grid Co.,Ltd.,Zhongshan 528401,Guangdong,China)

机构地区：[1]广东电网有限责任公司中山供电局,广东中山528401

出　　处：《电气传动》2023年第11期90-96,共7页Electric Drive

基　　金：广东电网有限责任公司科技项目(GDKJXM20184171)。

摘　　要：继保压板状态识别受各种环境因素影响长期依赖人工巡检,而目前流行的压板图像识别方案也依赖庞大后台计算架构,无法在操作现场即刻获取核对信息,严重影响工作效率。针对该问题,提出一种基于嵌入式的手持终端的解决方案。所提出的移动智能终端采用微型智能高清相机,实现了对压板屏柜多角度拍摄,有效地解决了由于玻璃门反光、异物阻挡、角度等妨碍拍摄的问题。所提出的基于嵌入式及YOLO Nano算法的继保压板状态现场识别的方法,选用高性能计算核心和专用视觉识别模块作为硬件平台,避免了采用庞大的后台计算架构,实现了操作现场的即时识别与核对。试验结果表明,该系统对继保压板状态识别的准确率达100%,有效地解决了继保压板状态的操作现场识别问题,具备较强的实用性和推广性。Due to various environmental factors,the relay platen status recognition needs manual inspection for a long time,and the current popular platen image recognition program requires a huge background computing architecture,which can not immediately obtain the verification information in the operation site,and seriously affect the work efficiency of the problem.To address this issue,a solution based on embedded handheld terminal was proposed.The proposed mobile intelligent terminal used a miniature intelligent HD camera to achieve multi-angle shooting of the platen screen cabinet,and effectively solved the problem of hindering shooting due to reflection of glass doors,foreign objects blocking and angles.The proposed relay platen status field recognition method based on embedded and you only look once(YOLO)Nano algorithm selected high performance computing core and special vision recognition module as hardware platform,which avoided huge background computing architecture and realized operation field real-time recognition and verification.The test results show that the system has 100%accuracy rate of relay platen status recognition,which effectively solves the operation site recognition problem of relay platen status and has strong practicality and popularization.

关 键 词：继电保护 继保压板状态现场识别 嵌入式图像识别系统 手持终端 YOLO Nano算法 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]