Moving cast shadow detection using block nonnegative matrix factorization

X. Yang, D. Liu, D. Zhou, R. Yang

Research output: Contribution to journalArticlepeer-review

3 Scopus citations

Abstract

In recent years, moving cast shadow detection has become a critical challenge in improving the accuracy of moving object detection in video surveillance. In this paper, we propose two novel moving cast shadow detection methods based on nonnegative matrix factorization (NMF) and block nonnegative matrix factorization (BNMF). First, the algorithm of moving cast shadow detection using NMF is given and the key points such as the determination of moving shadow areas and the choice of discriminant function are specified. Then BNMF are introduced so that the new training samples and new classes can be added constantly with lower computational complexity. Finally, the improved shadow detection method is detailed described according to BNMF. The effectiveness of proposed methods is evaluated in various scenes. Experimental results demonstrate that the method achieves high detection rate and outperforms several state-of-the-art methods.

Original languageEnglish
Pages (from-to)229-234
Number of pages6
JournalBulletin of the Polish Academy of Sciences: Technical Sciences
Volume66
Issue number2
DOIs
StatePublished - Apr 2018

Bibliographical note

Funding Information:
Acknowledgements. This research was supported by the National Natural Science Foundation of China (61573182) and by the Fundamental Research Funds for the Central Universities (NS2014035).

Publisher Copyright:
© 2018 De Gruyter Open Ltd. All rights reserved.

Keywords

  • Block nonnegative matrix factorization
  • Moving cast shadow detection
  • Nonnegative matrix factorization
  • Video surveillance

ASJC Scopus subject areas

  • Information Systems
  • Atomic and Molecular Physics, and Optics
  • Engineering (all)
  • Computer Networks and Communications
  • Artificial Intelligence

Fingerprint

Dive into the research topics of 'Moving cast shadow detection using block nonnegative matrix factorization'. Together they form a unique fingerprint.

Cite this