摘要
架空导线检测是电力巡检中的重要内容。传统的架空导线检测方法可以较好地识别架空导线的形态特征,但是其场景鲁棒性有限,难以排除周围环境干扰。基于深度语义分割的架空导线检测是近年来新兴的检测方法,其场景鲁棒性较好,但在实践中,此类方法的检测结果存在分割区域不连续的问题。针对以上问题,提出基于交叉检验的架空导线检测方法。首先采用马尔可夫链边缘线段检测器(Markov chain marginal line segment detector,MCMLSD)和DeepLabv3+分割网络分别对架空导线进行检测,然后设计交叉检验算法对上述2种方法的检测结果进行比较和融合,成功提取出完整的架空导线并抑制了周围环境中线段结构对检测结果的影响。实验证明该方法在天空、土地、树林等场景中均具有较好的鲁棒性,为无人机架空导线自动巡检提供了新思路,推进了该领域的发展。
Detection of overhead transmission line is an important part of power inspection.The traditional detection method can identify the morphological characteristics of overhead transmission line,but its scene robustness is limited and it is difficult to exclude the interference from the surrounding environment.The method of overhead transmission line inspection based on the semantic segmentation network is a new detection method which has better robustness to different scenes.However,in practice,the result of this method may have the discontinuity of the detected area of overhead transmission line.To solve the above problems,this paper presents an method of overhead transmission line detection based on cross-check.First,MCMLSD algorithm(Markov Chain Marginal Line Segment Detector) and DeepLabv3+ semantic segmentation network are used to detect overhead transmission line respectively,and then a cross-check algorithm is designed to compare and fuse the detection results of these two methods.The complete overhead transmission lines can be successfully extracted and the influence of the segment structures in the surrounding environment on the detection results is suppressed.Experiments show that this method is robust to the scenes of sky,land,and forest.It provides a new idea for the automatic inspection of transmission line inspection and promotes the development of this field.
作者
高超
吴奇伟
杨凡
乔家齐
GAO Chao;WU Qiwei;YANG Fan;QIAO Jiaqi(State Grid Jiangsu Electric Power Co.,Ltd,Nanjing,Jiangsu 210004,China;School of Electric Power Engineering,Nanjing Institute of Technology,Nanjing,Jiangsu 211167,China)
出处
《广东电力》
2023年第8期124-130,共7页
Guangdong Electric Power
基金
江苏省社会发展重点研发基金项目(BE2022789)
南京工程学院科研项目(ZKJ202003,科22-318)。
作者简介
高超(1983),男,河南信阳人,高级工程师,硕士,研究方向为输电线路智能运检管理及输电数字化技术应用,E-mail:13566617@qq.com;吴奇伟(1992),男,江苏盐城人,工程师,研究方向为输电运检及数字化建设,E-mail:972538297@qq.com;杨凡(2000),男,安徽合肥人,在读硕士研究生,研究方向为无人机架空导线实时提取,E-mail:1931394825@qq.com;乔家齐(1997),男,河南新乡人,在读硕士研究生,研究方向为无人机架空导线智能定位与管理,E-mail:1061377627@qq.com。
