EpNet : Power lines foreign object detection with Edge Proposal Network and data composition

Research output: Journal Publications and Reviews (RGC: 21, 22, 62)21_Publication in refereed journalpeer-review

View graph of relations

Author(s)

  • Junyu Su
  • Yukun Su
  • Yu Zhang
  • Weiqiang Yang
  • Huichou Huang
  • And 1 others
  • Qingyao Wu

Detail(s)

Original languageEnglish
Article number108857
Journal / PublicationKnowledge-Based Systems
Volume249
Online published27 Apr 2022
Publication statusPublished - 5 Aug 2022
Externally publishedYes

Abstract

Power lines foreign object detection task is to detect objects suspending on power lines, they might be kites, plastic bags or anything else, which could be a potential risk to power system. However, this task remains a challenge due to the lack of data, because these data can only be produced by a few major video surveillance companies, who treat their data as valuable property and will not share it with others. Without massive training data, we could not obtain an excellent neural network. In this paper, we introduce a new data composition method to generate artificial data and help alleviate the data shortage problem. What is more, we propose a new detection method called Edge Proposal Network (EpNet) to reduce wrong proposal locations and increase detection performance. At last, we conduct several experiments to verify the effectiveness of the two methods, and some discussion experiments to gain a deeper understanding of the composited data.

Research Area(s)

  • Data composition, Foreign object detection, Power lines

Citation Format(s)

EpNet : Power lines foreign object detection with Edge Proposal Network and data composition. / Su, Junyu; Su, Yukun; Zhang, Yu; Yang, Weiqiang; Huang, Huichou; Wu, Qingyao.

In: Knowledge-Based Systems, Vol. 249, 108857, 05.08.2022.

Research output: Journal Publications and Reviews (RGC: 21, 22, 62)21_Publication in refereed journalpeer-review