基于上下文信息增强与特征细化的绝缘子破损检测方法

Detection Method of Insulator Breakage Based on Context Augmentation and Feature Refinement

快速准确地检测出绝缘子缺陷是电网维护的重要任务,也极具挑战性。针对目前主流绝缘子缺陷检测算法检测速度慢且模型复杂度较高的问题,提出一种基于上下文信息增强与特征细化的绝缘子破损检测方法。该方法采用轻量化的ECA-Ghost Net作为骨干网络,骨干网络输出端嵌入轻量化的自适应上下文信息增强模块,为绝缘子破损缺陷注入多尺度上下文信息;然后在特征金字塔输出端引入快速且高效的特征细化模块,用于增强绝缘子破损缺陷特征。在该文构建的数据集上进行了多组对比实验,结果表明该文提出的方法均值平均精度可达约97.05%,检测速度约为63帧/s,模型计算量和参数量分别为1.46 G和1.68 M,各项性能指标均优于Retina Net、YOLOv4和YOLOF等主流算法。该文研究结果可为无人机嵌入式应用提供参考。

Rapid and accurate detection of insulator defects is an important and challenging task for grid maintenance.Aiming at the problems of slow detection speed and high model complexity of the current mainstream insulator defect detection algorithms,we propose a breakage detection method of insulator based on context augmentation and feature re-finement.In the method,firstly,the lightweight ECA-GhostNet is used as the backbone,and a lightweight adaptive context augmentation module is embedded at the output of the backbone to inject multi-scale context for the insulator breakage.Then,a fast and efficient feature refinement module are introduced at the output of the feature pyramid net-work(FPN)for augmenting insulator breakage defect features.Several sets of comparative experiments are carried out on the dataset constructed in this paper.The results show that the mean average precision(mAP)of the method proposed in this paper can reach about 97.05%,the detection speed is about 63 Frame/s,and the floating point operations(FLOPs)and parameters are 1.46 G and 1.68 M,respectively.All performance indicators are superior to those obtained by mainstream algorithms,such as RetinaNet,YOLOv4 and YOLOF.The research results can provide a reference for embedded applica-tions of drone.