应用YOLOv4-tiny算法实现保护压板智能校核  

Application of YOLOv4-tiny Algorithm for Intelligent Calibration of Protection Platens

作　　者：杨宗源 侯进[1,2] 周浩然 郝彦超 文志龙 李天宇 YANG Zong-yuan;HOU Jin;ZHOU Hao-ran;HAO Yan-chao;WEN Zhi-long;LI Tian-yu(School of Information Science and Technology, Southwest Jiaotong University, Chengdu 611756, China;Southwest Jiaotong University National Engineering Laboratory of Integrated Transportation Big Data Application Technology, Chengdu 611756, China;Graduate School of Tangshan, Southwest Jiaotong University, Tangshan 063000, China)

机构地区：[1]西南交通大学信息科学与技术学院,成都611756 [2]西南交通大学综合交通大数据应用技术国家工程实验室,成都611756 [3]西南交通大学唐山研究生院,唐山063000

出　　处：《科学技术与工程》2022年第2期570-576,共7页Science Technology and Engineering

基　　金：国家重点基础研究发展计划(2014CB845800);四川省科技计划(2020SYSY0016)。

摘　　要：目前继电保护压板的巡检校核仍以人工为主,为提高其工作的效率,提出了一种智能实时校核方法。该方法首先使用YOLOv4-tiny算法对压板的投退状态进行预测,然后使用腾讯开源的ncnn前向推理框架,对YOLO模型进行优化,最后将模型移植到移动端,使用手机软件完成压板校核。经测试,模型的均值平均精度达到99.13%,平均预测速度达到每秒30张图片,并可以有效解决反光、遮挡等环境因素的影响,可以显著提升巡检工作的效率。At present,the inspection and calibration of relay protection platens is still mainly manual.To improve the efficiency of its work,an intelligent real-time verification method was proposed.First,the YOLOv4-tiny algorithm was used to predict the throwback state of the pressure plate.Then,Tencent's open-source ncnn forward inference framework was used to optimize the YOLO model.Finally,the model was transplanted to the mobile terminal to complete the platens calibration by using the cell phone software.After testing,the mean average accuracy of the model reaches 99.13%,the average prediction speed reaches 30 images per second,and can effectively solve the influence of environmental factors such as reflection and shading,which can significantly improve the efficiency of inspection work.

关 键 词：保护压板 智能校核 YOLOv4-tiny ncnn模型 移动端 

分 类 号：TM769[电气工程—电力系统及自动化]