基于Flexible YOLOv7的输电线路绝缘子缺陷检测和故障预警方法

Insulator Defect Detection and Fault Warning Method for Transmission Line Based on Flexible YOLOv7

电力设备的平稳运行是保障居民生产生活的重要前提。输电线路绝缘子缺陷尺寸较小,传统的目标检测算法通常难以识别到缺陷目标,误检、漏检率较高。针对不同材质绝缘子缺陷检测存在目标过小、遮挡、背景复杂等难题,提出了一种基于Flexible YOLOv7的绝缘子缺陷检测算法。该算法继承了YOLOv7网络的E-ELAN结构、Rep重参数化和辅助训练策略,并且在特征提取的过程中集成GAM注意力机制以放大显著的跨维度接受区域,通过高效的Ghost SPPCSPC结构减少模型训练过程中的参数冗余,引入Efficient IOU Loss重点关注高质量的anchors提升原始模型的检测精度。最后通过图像后处理技术对绝缘子缺陷进行等级划分与精细计算,并结合算法部署开发了绝缘子缺陷故障检测系统用于故障的提前预警。实验结果表明,该算法在密集目标、遮挡、小目标缺陷检测中的平均准确率AP、召回率Recall、相关指标F1指标均领先于当前先进的几类目标检测算法,在复杂环境下的绝缘子缺陷检测和故障预警方面具有一定的现实意义。

The stable operation of power equipment is an important prerequisite for ensuring production and life.The defect size of insulator on transmission line is small.The traditional target detection algorithm is usually difficult to iden-tify the defect target,and the false detection and missed detection rate are high.Therefore,an insulator defect detection algorithm based on Flexible YOLOv7 is proposed to solve the problems of small target,occlusion and complex back-ground in defect detection of insulators with different materials.The algorithm inherits the E-ELAN structure,Rep re-parameterization and auxiliary training strategy of YOLOv7 network,and integrates the GAM attention mechanism in the process of feature extraction to enlarge the significant cross-dimensional acceptance region.The efficient Ghost SPPCSPC structure reduces the parameter redundancy in the model training process,and introduces Efficient IOU Loss to focus on the high-quality anchor box to improve the detection accuracy of the model.Finally,the insulator defects are graded and finely calculated by image post-processing technology,and an insulator defect fault detection system is de-veloped for early warning of faults in combination with algorithm deployment.The experimental results show that the AP,Recall and F1 indexes of the proposed algorithm in the two types of defects of dense target,occlusion and small target de-tection are all ahead of the current advanced target detection algorithms.The method has certain practical significance in insulator defect detection and fault warning in complex environments.