基于YOLOv3的绝缘子串爆裂缺陷检测研究

Research on detection method of insulator chain defects based on YOLOv3

针对绝缘子爆裂缺陷检测对时效性的要求,提出了一种基于YOLOv3的绝缘子缺陷检测模型。该模型针对绝缘子爆裂缺陷检测的要求,在YOLOv3算法的基础上进行改进。为提高模型检测速度,将YOLOv3的主干特征提取网络替换为更轻量的GhostNet网络;为提升模型精度,采用K-means++进行聚类以改进模型的先验框尺寸,并将回归损失函数改进为更能体现检测效果的CIoU损失函数。实验结果表明,该模型的mAP为90.32%,相比原YOLOv3提升了3.65%,能够更高效地识别绝缘子及其爆裂缺陷。

Based on the timeliness requirements of insulator burst defect detection,a YOLOv3-based insulator defect detection model is proposed.The model is improved on the basis of the YOLOv3 algorithm to meet the requirements of insulator burst defect detection.In order to improve the speed of model detection,the backbone feature extraction network of YOLOv3 is replaced with a lighter network which is called GhostNet.At the same time,to improve the accuracy of the model,K-means++is used for clustering to make the prior frame size of the model more accurate,and the regression loss function is improved to a CIoU function that can reflect the detection effect better.The experiment results show that the mAP of the model is 90.32%,which is 3.65%higher than the original YOLOv3,and the model can identify insulators and their bursting defects more efficiently.