The detection of defects in ground wires is often slowed down and less accurate due to the large size of UAV aerial images and complex background environment.To address this issue,this paper proposes a lightweight ground wire defect detection method based on LSD and deep learning.First,the LSD algorithm is used to extract linear features from the images.Then,a segmentation baseline is fit by combining it with RANSAC.Based on the segmentation baseline,the region of the ground wire is now clearly segmented,background interferences are eliminated,and image size sent to the detection network reduced.After modifying the YOLOv5 backbone network,the number of parameters is reduced,making it easier to deploy on edge computing equipment.The proposed method reduces the inspected area and suppresses interference.It improves the accuracy from the original 67.9％to 71.3％,at the same time,the detection speed is increased by 12％.It has the advantages of high precision and fast speed,which is suitable for deployment on edge computing equipment.
关键词
导地线缺陷检测/航拍图像/直线段检测/直线拟合/深度学习
Key words
detection of defects in ground wires/UAV aerial images/line segment detect/line fitting/artificial intelligence
维普
万方数据