Infrared dim and small target detection based on spatio-temporal spectral saliency

Kai Zhang, Chenhui Li, Shaoyi Li, Xiaotian Wang, Saisai Niu

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

Abstract

Infrared dim and small target detection plays a critical role in space-based infrared system and infrared search and track applications. Inspired by the visual saliency of human beings, the present study proposes an infrared dim and small target detection method based on the spatio-temporal and spectral saliency. In this regards, the 2D-DoG filter is initially applied to preprocess the image. Then, four data channels of quaternion are constructed through contrast measurement, velocity estimation and the direction feature extraction so that the saliency detection in the frequency domain is completed. Finally, considering the motion characteristics of the target, the false alarm rate is reduced by multi-frame confirmation of the pipeline filtering. Experimental results show that the proposed method can achieve stable and continuous target detection for different backgrounds.

Original languageEnglish
Title of host publicationFuzzy Systems and Data Mining V - Proceedings of FSDM 2019
EditorsAntonio J. Tallon-Ballesteros
PublisherIOS Press BV
Pages1118-1123
Number of pages6
ISBN (Electronic)9781643680187
DOIs
StatePublished - 2 Oct 2019
Event5th International Conference on Fuzzy Systems and Data Mining, FSDM 2019 - Kitakyushu, Japan
Duration: 18 Oct 201921 Oct 2019

Publication series

NameFrontiers in Artificial Intelligence and Applications
Volume320
ISSN (Print)0922-6389
ISSN (Electronic)1879-8314

Conference

Conference5th International Conference on Fuzzy Systems and Data Mining, FSDM 2019
Country/TerritoryJapan
CityKitakyushu
Period18/10/1921/10/19

Keywords

  • Human visual system
  • Infrared dim and small target detection
  • Pipeline filtering
  • Signal-clutter ratio
  • Spatio-temporal spectral saliency

Fingerprint

Dive into the research topics of 'Infrared dim and small target detection based on spatio-temporal spectral saliency'. Together they form a unique fingerprint.

Cite this