Journal of South China University of Technology(Natural Science Edition) ›› 2019, Vol. 47 ›› Issue (2): 77-84.doi: 10.12141/j.issn.1000-565X.180295

• Computer Science & Technology • Previous Articles     Next Articles

Adaptive Dual Fractional-order Optical Flow Model for Motion Segmentation
 

 ZHU Bin1 TIAN Lianfang1， 2， 3 DU Qiliang1， 2， 3 YU Lubin1    

  1.  1． School of Automation Science and Engineering，South China University of Technology，Guangzhou 510640，Guangdong，China; 2． Ｒesearch Institute of Modern Industrial Innovation，South China University of Technology，Zhuhai 519000，Guangdong， China; 3． Key Laboratory of Autonomous Systems and Network Control of Ministry of Education， Guangzhou 510640，Guangdong，China
  • Received:2018-06-11 Revised:2018-07-17 Online:2019-02-25 Published:2019-01-02
  • Contact: 杜启亮( 1980) ，男，副研究员，主要从事模式识别与机器视觉研究 E-mail:qldu@scut.edu.cn
  • About author:朱斌( 1982) ，男，博士生，主要从事模式识别与机器视觉研究
  • Supported by:
    Supported by the Frontier and Key Technological Innovation Special Fund of Guangdong Province(2016B090912001) 

PDF

77

Abstract: An adaptive dual fractional order variational optical flow ( ADFOVOF) model was proposed to solve the problem that the current optical flow algorithm cannot segment moving objects in weak texture region under insufficient illumination． The model was an adaption version of dual fractional order optical flow ( DFOVOF) model that adopts fractional differential mask in both，the data term and smoothness term of traditional Horn ＆ Schunck ( HS) model． Specially，the order and size of fractional order differential mask for each region were adjusted by image signal to noise ratio ( SNＲ) whereas，the shape of the fractional order differential mask was regulated by super-pixel that prevents the mask to interfere by surrounding regions． The experiment results show that our algorithm can get accurate optical flow field and obtain clear motion outlines under insufficient illumination or illumination changes scenes． 

Key words: image sequence analysis, variational optical flow model, fraction order differential mask, motion detection

CLC Number: