基于改进YOLOV4-tiny的电力设备红外识别研究  

Research on infrared recognition of power equipment based on improved YOLOV4-tiny

作　　者：刘冲冲 张力平[1,2] 任锦飞 陈群元 LIU Chongchong;ZHANG Liping;REN Jinfei;CHEN Qunyuan

机构地区：[1]长安大学工程机械学院,陕西西安710064 [2]长安大学公路养护装备国家工程实验室,陕西西安710068 [3]不详

出　　处：《现代机械》2024年第1期83-87,共5页Modern Machinery

摘　　要：红外检测技术因其具有无需直接接触、带电状态进行检测、检测快速等优点,广泛应用于电力设备过热故障诊断领域。但是,红外图像质量不佳和电力设备的复杂分布等问题,对电力设备故障红外检测过程中的电力设备种类检测精度带来不利影响。为了实现电力设备种类识别的快速准确检测,在YOLOV4-tiny目标检测算法的基础上,提出了适用于电力设备过热故障诊断过程设备种类识别的YOLOV4-tiny目标检测模型。通过旋转矩形框机制替换水平矩形框机制、改进激活函数和采用PAN+FPN加强特征提取网络的方式,来改进原检测模型,使其在检测时更精准更快速。通过多次实验验证,优化后的模型相较于YOLOV4、YOLOV4-tiny两种模型来说,检测速度没有明显提升,但其检测精度提升了1.89%,这为电力设备过热故障红外诊断过程中的设备种类识别研究带来了新的思路。Infrared detection technology is widely used in the field of overheat fault diagnosis of power equipment because of its advantages such as no direct contact,live detection and fast detection.However,the poor infrared image quality and the complex distribution of power equipment have an adverse effect on the detection accuracy of power equipment type in the process of power equipment fault detection.In order to realize the fast and accurate identification of power equipment type,this paper proposes a YOLOV4-tiny object detection model suitable for the identification of power equipment type in the process of overheat fault diagnosis based on the YOLOV4-tiny object detection algorithm.The original detection model is improved by replacing the horizontal rectangular frame mechanism with the rotating rectangular frame mechanism,improving the activation function and using PAN+FPN to strengthen the feature extraction network,so as to make the detection more accurate and faster.Through many experiments,the optimized model has no significant improvement in detection speed compared with YOLOV4 and YOLOV4-TINY models,but its detection accuracy is increased by 1.89%.This study has brought a new idea for the equipment type identification in the infrared diagnosis process of power equipment overheating fault.

关 键 词：红外检测 带电检测 旋转矩形 YOLOV4-tiny 

分 类 号：TM711[电气工程—电力系统及自动化]