Real-time vehicle tracking using convolutional neural networks in aerial video

Yu Yang, Chengpo Mu*, Ruixin Yang, Yanjie Wang

*Corresponding author for this work

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

1 Citation (Scopus)

Abstract

Vehicle tracking based on video images has been widely used in military and civilian fields. The tracking method must robust enough to hand the unexpected situations that may occur during the tracking process. In this paper, a novel vehicle tracking method based on convolutional neural networks (CNNs) is proposed to target the accurate and speed demand of vehicle tracking. The proposed method contains two networks with shared weights. It utilizes the residual block to reduce the train error. Offline training is used to achieve real-time tracking. It also use transfer learning to reduce training time. The experimental results under the real aerial video demonstrate that vehicle tracker achieves an accuracy of 70.8% and the speed of 135fps with GPU. The proposed method is robust enough to handle occlusion and other interference conditions.

Original languageEnglish
Title of host publicationProceedings - 2019 International Conference on Sensing, Diagnostics, Prognostics, and Control, SDPC 2019
EditorsChuan Li, Shaohui Zhang, Jianyu Long, Diego Cabrera, Ping Ding
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages251-255
Number of pages5
ISBN (Electronic)9781728101996
DOIs
Publication statusPublished - Aug 2019
Event2019 International Conference on Sensing, Diagnostics, Prognostics, and Control, SDPC 2019 - Beijing, China
Duration: 15 Aug 201917 Aug 2019

Publication series

NameProceedings - 2019 International Conference on Sensing, Diagnostics, Prognostics, and Control, SDPC 2019

Conference

Conference2019 International Conference on Sensing, Diagnostics, Prognostics, and Control, SDPC 2019
Country/TerritoryChina
CityBeijing
Period15/08/1917/08/19

Keywords

  • Convolutional neural networks
  • Real-time
  • Residual Network
  • Vehicle tracking

Fingerprint

Dive into the research topics of 'Real-time vehicle tracking using convolutional neural networks in aerial video'. Together they form a unique fingerprint.

Cite this