基于无监督知识蒸馏的高铁绝缘子缺陷检测算法

An Algorithm for Detection of Insulator Defects for High-speed Railway based on Unsupervised Knowledge Distillation

绝缘子作为高铁接触网的重要组成部分,是隔离接触网和地面的关键零部件。但绝缘子缺陷样本稀缺,缺陷类型多样,基于有监督深度学习的绝缘子缺陷检测难以满足现场需求。针对该问题,本文提出一种基于知识蒸馏的绝缘子缺陷检测算法。该方法主要分为两个阶段:第一阶段使用一个基于YOLO-v5的绝缘子旋转定位网络快速准确地将绝缘子从接触网图像中提取出来,并使用滑动窗口将其下采样为绝缘子小块;第二阶段使用由一个教师网络和一个学生网络组成的绝缘子缺陷检测网络检测缺陷。实验表明该方法具有良好的可行性。

As an important part of OCL of high-speed railway,insulator is a key component to isolate the OCL from the ground.However,due to the scarcity of insulator defect samples and the variety of defect types,the insulator defect detection based on supervised in-depth learning is difficult to meet the on-site demand.To solve this problem,the paper proposes an insulator defect detection algorithm based on knowledge distillation.The method is mainly divided into two stages:in the first stage,an insulator rotation positioning network based on YOLO-v5 is used to quickly and accurately extract the insulator from the OCL image,and the sliding window is used to sample it into insulator blocks;in the second stage,insulator defect detection network consisting of a teacher network and a student network is used to detect the defects.The experiment shows that the method is feasible.