基于改进YOLOv5s的高压输电线路巡检故障实时检测方法

Real Time Fault Detection Method for High-voltage Transmission Line Inspection Based on Improved YOLOv5 s

文章针对传统的高压输电线路巡检过程中存在的效率低下等问题,提出一种基于改进YOLOv5s的高压输电线路巡检故障实时检测方法。将卷积块注意力模块加入YOLOv5s骨干网络中最后一层,从通道注意力与空间注意力两个方面对特征图进行权重分配。所提出的方法在无人机获取的图像中进行训练和验证,结果表明所提出的方法相较基于YOLOv3、YOLOv5s,能保持较高检测精度的同时,还能保持较快的速度。

A real-time fault detection method for high-voltage transmission line inspection based on improved YOLOv5 s is proposed to address the issues of low efficiency in traditional inspection processes.Add the Convolutional Block Attention Module(CBAM)to the last layer of the YOLOv5 s backbone network,and assign weights to the feature maps from two aspects:channel attention and spatial attention.The proposed method was trained and validated in images obtained by drones,and the results showed that compared to YOLOv3 and YOLOv5 s,the proposed method can maintain high detection accuracy while maintaining faster speed.