基于目标检测与特征匹配测距的输电线路防外破监测方法
A Method of Transmission Line External Break-proof Monitoring Based on Target Detection and Feature Matching Ranging
佘仰腾 1范亮 2张惠荣 1陈奕戈 1张磊2
作者信息
- 1. 广东电网有限责任公司惠州供电局 广东 惠州 516000
- 2. 广州中科智巡科技有限公司 广东 广州 510623
- 折叠
摘要
本文基于双目摄像机,提出一种结合目标检测与特征匹配测距的输电线路外力破坏隐患检测方法,首先通过改进YOLOv4模型检测出图像中的隐患目标区域;再利用LoFTR模型对双目摄像机所拍摄左右图像中的隐患目标区域进行特征匹配,计算区域中若干特征匹配对的像素差均值对隐患目标进行测距.最终将隐患目标距离大于阈值的视为无关目标,排除误检,而对距离小于阈值的目标,则通过deepsort目标跟踪算法计算停留帧数,根据停留时长进行告警判读,实现自动化、高准确率、高实用性的输电线路防外破监测.
Abstract
Based on binocular cameras,this paper proposes a detection method for hidden danger of external force damage of transmission line combining target detection and feature matching ranging.Firstly,the hidden danger target area in the image is detected by improving YOLOv4 model;Then,the LoFTR model is used to match the features of the hidden danger target areas in the left and right images taken by the binocular camera,and calculate the average pixel difference of several feature matching pairs in the area to measure the distance of the hidden danger target.Finally,the hidden danger target whose distance is greater than the threshold value is regarded as irrelevant target,and the false detection is eliminated.For the target whose distance is less than the threshold value,the number of dwell frames is calculated by the deepsort target tracking algorithm,and the alarm interpretation is performed according to the dwell time,so as to realize the automation,high accuracy,and high practicability of transmission line external break-proof monitoring.
关键词
双目图像处理/LoFTR/deepsort目标跟踪算法Key words
binocular image processing/LoFTR/deepsort target tracking algorithm引用本文复制引用
基金项目
中国南方电网有限责任公司科技项目(031300KK52220013)
出版年
2024
NETL
NSTL
万方数据