Skip to main navigation Skip to search Skip to main content

Weighted Pyramid Network for Foreign Object Debris Detection on Airfield Pavement

  • Maji Huang
  • , Xiaoguang Cao
  • , Xiangzhi Bai*
  • , Yufeng Gu
  • *Corresponding author for this work
  • Beihang University

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

In this paper, a novel network architecture Weighted Pyramid Network (WPN) is proposed to detect foreign object debris (FOD) on airfield pavement. WPN is a region-based algorithm and it produces region proposals from multi-scale layers with different ratio. Fed to fully-connected layer, these multi-scale proposals are exacted to the same size by undergoing a RoIAlign layer. Moreover, Finetune Loss is designed in WPN to concentrate the model on hard examples and this practice improves the efficiency of training a lot. In the end, compared to some popular object detection network, WPN is proved to achieve better result for FOD detection on airfield pavement.

Original languageEnglish
Title of host publication2018 14th IEEE International Conference on Signal Processing Proceedings, ICSP 2018
EditorsYuan Baozong, Ruan Qiuqi, Zhao Yao, An Gaoyun
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages625-628
Number of pages4
ISBN (Electronic)9781538646724
DOIs
StatePublished - 2 Feb 2019
Event14th IEEE International Conference on Signal Processing, ICSP 2018 - Beijing, China
Duration: 12 Aug 201816 Aug 2018

Publication series

NameInternational Conference on Signal Processing Proceedings, ICSP
Volume2018-August

Conference

Conference14th IEEE International Conference on Signal Processing, ICSP 2018
Country/TerritoryChina
CityBeijing
Period12/08/1816/08/18

Keywords

  • Finetune Loss
  • Foreign Object Debris
  • Object Detection
  • RoIAlign
  • Weighted Pyramid Network

Fingerprint

Dive into the research topics of 'Weighted Pyramid Network for Foreign Object Debris Detection on Airfield Pavement'. Together they form a unique fingerprint.

Cite this