计算机工程与应用 ›› 2018, Vol. 54 ›› Issue (15): 1-5.DOI: 10.3778/j.issn.1002-8331.1708-0402

• 热点与综述 • 上一篇    下一篇

基于超像素和判别稀疏的运动目标跟踪算法

邱晓荣1,2,彭  力3,刘全胜2,3   

  1. 1.马来西亚管理科学大学 信息科学与工程学院,雪兰莪 莎阿南 40100
    2.无锡职业技术学院 物联网技术学院,江苏 无锡 214121
    3.江南大学 物联网工程学院,江苏 无锡 214122
  • 出版日期:2018-08-01 发布日期:2018-07-26

Object tracking algorithm based on superpixel and discriminative sparsity

QIU Xiaorong1,2, PENG Li3, LIU Quansheng2,3   

  1. 1.Faculty of Information Sciences & Engineering, Management & Science University, Shah Alam 40100, Malaysia
    2.School of Internet of Things Technology, Wuxi Institute of Technology, Wuxi, Jiangsu 214121, China
    3.School of Internet of Things Engineering, Jiangnan University, Wuxi, Jiangsu 214122, China
  • Online:2018-08-01 Published:2018-07-26

摘要: 针对目标遮挡、非刚性变换、光照变换等因素干扰产生的漂移问题,提出基于超像素和判别稀疏的运动目标跟踪算法。算法首先利用SLIC方法对运动目标的观测区域进行超像素分割,然后通过K-Means算法构建包含目标和背景的超像素字典,再基于判别稀疏表示和[?1]范数最小化框架求解候选目标的稀疏系数,同时结合粒子滤波框架和在线字典更新策略完成目标跟踪。实验结果表明,该算法在多种因素干扰的环境中具有较强的鲁棒性,能够准确稳定地进行在线目标跟踪。

关键词: 判别稀疏, 超像素, 目标跟踪, 表观模型

Abstract: According to the drift problems caused by occlusion, deformation and illumination, an algorithm based on superpixel and discriminative sparsity is proposed. Firstly, the object observation area is segmented into superpixels by using SLIC method. Secondly, K-Means algorithm is adopted to generate superpixel dictionary combined with target and background information. Then, sparse coefficients of candidate targets are solved based on discriminative sparsity representation and [?1]-norm minimization framework. Finally, particle filter framework and online dictionary updating are applied to accomplish object tracking. Both quantitative and qualitative experimental results show the proposed algorithm performs more robustly under the environment of multiple interference factors, and it also can perform online object tracking more accurately and steadily.

Key words: discriminative sparsity, superpixel, object tracking, representation model