摘要
为了解决目前利用图像识别技术检测电力线搭挂异物所存在的鲁棒性低、准确性低、缺少精确空间定位信息等问题,提出一种基于激光点云数据的电力线搭挂异物检测及精准定位方法。首先,从正常未搭挂异物电力线点云线性特征和平滑性特征着手,分析电力线异物在激光点云数据中空间特性;然后,提取电力线点云水平和垂直方向三维模型残差,通过无监督残差聚类实现单根电力线分割;之后,对提取的单根电力线网格化,根据异物处网格突变的空间特性粗提取异物区域;最后,提出半空心球邻域的定义,基于半空心球邻域内角度过滤方法精确定位异物点云。实验结果表明,对于不同情况下异物,均能准确检测并精确定位,平均检测准确率100%,平均定位精确率97.75%,平均运行时间0.175秒,同时可实现异物搭挂在地线还是导线的准确判断,与图像识别方法对比,该方法不存在拍摄角度、拍摄环境、训练样本量的影响,鲁棒性高,使用范围广,工程实用前景广阔。
In order to solve the problems of low robustness,low accuracy and lack of accurate spatial location information in the detection of power line hanging foreign bodies by using image recognition technology,in this paper,a method of power line hanging foreign body detection and accurate location based on laser point cloud data is proposed.First of all,starting from the linear and smoothness characteristics of the normal unmounted power line point cloud,the spatial characteristics of the power line foreign body in the laser point cloud data are analyzed,and then the horizontal and vertical three-dimensional model residuals of the power line point cloud are extracted.The clustering algorithm is used to classify the residual,so as to realize the segmentation of a single power line.After that,the extracted single power line is meshed and determined,and the region of the foreign body is roughly extracted according to the sudden change of the grid at the foreign body.finally,the definition of semi-hollow sphere neighborhood is proposed.the foreign body point cloud is located accurately based on the angle filtering method in the neighborhood of the semi-hollow sphere.The experimental results show that the foreign body can be accurately detected and located under different conditions,with an average accuracy of 100%,an average accuracy of 97.75%,and an average running time of 0.175 seconds,compared with the image recognition method,this method does not have the influence of shooting angle,shooting environment and training sample size,and has high robustness and wide application range,the engineering application prospect is broad.
作者
虢韬
徐梁刚
史洪云
王迪
龙贤哲
龙新
GUO Tao;XU Lianggang;SHI Hongyun;WANG Di;LONG Xianzhe;LONG Xin(Production and Technology Department of Guizhou Power Grid Co.,Ltd.,Guiyang 550000 Guizhou,China;Transmission Operation and Maintenance Branch of Guizhou Power Grid Co.,Ltd.,Guiyang 550000 Guizhou China;China Power Construction Group Guizhou Electric Power Design&Research Institute Co.,Ltd.,Guiyang 550000 Guizhou,China)
出处
《电力大数据》
2021年第7期9-16,共8页
Power Systems and Big Data
关键词
电力线异物检测
激光点云
空间特性
网格突变
半空心球邻域
角度过滤
power line foreign body detection
laser point cloud
spatial characteristics
sudden change of grid
semi-hollow sphere neighborhood
angle filtering
