山东电力技术2025，Vol.52Issue(5)：9-17,9.DOI:10.20097/j.cnki.issn1007-9904.2025.05.002
结合对比损失和并联残差注意力的绝缘子多缺陷检测
Multi-Defect Detection of Insulators Combining Contrastive Loss and Parallel Residual Attention
摘要
Abstract
Accurate detection of insulator defects in transmission lines is crucial for ensuring the safety and stability of power transmission.To address the challenges posed by complex and variable backgrounds,as well as the diverse forms of defects in aerial images for multi-defect detection of insulators,this paper proposes a method combining contrastive loss and parallel residual attention for insulator multi-defect detection.Specifically,the parallel residual attention mechanism reduces the mutual interference between channel and spatial attention,while introducing a residual structure to enhance the model's focus on defect areas,effectively addressing complex background issues while retaining key information.Contrastive loss maximizes the differences between features of different defects and minimizes the differences between features of the same defect,improving the model's ability to discriminate and express feature consistency for multi-form defects.In addition,the combination of focal shape Intersection over Union(IoU)and shape IoU optimizes both the shape and size of bounding boxes,as well as focus on high-quality bounding boxes,further enhancing bounding box regression accuracy.Experimental results show that the proposed method outperforms current mainstream algorithms.Compared to the baseline model YOLOv8,the accuracy of defect detection for insulator drop,damage,and flashover improved by 0.9%,6.0%,and 7.3%,respectively.This provides reliable technical support for power emergency monitoring and analysis,thus more effectively ensuring the safety and stability of power transmission.关键词
绝缘子缺陷/并联残差注意力/对比损失/应急监测/边界框回归Key words
insulator defects/parallel residual attention/contrastive loss/emergency monitoring/bounding box regression分类
计算机与自动化引用本文复制引用
肖甜甜,杨珂,王乾铭,翟永杰..结合对比损失和并联残差注意力的绝缘子多缺陷检测[J].山东电力技术,2025,52(5):9-17,9.基金项目
国家自然科学基金项目(62373151) （62373151）
河北省自然科学基金项目(F2023502010).National Natural Science Foundation of China(62373151) （F2023502010）
Natural Science Foundation of Hebei Province(F2023502010). （F2023502010）
