基于超像素和局部稀疏表示的目标跟踪方法

系统仿真学报 ›› 2016, Vol. 28 ›› Issue (5): 1017-1030.

基于超像素和局部稀疏表示的目标跟踪方法

杨恢先, 刘昭, 刘阳, 刘凡, 贺迪龙

湘潭大学物理光电工程学院,湖南湘潭 411105

收稿日期:2015-03-10 修回日期:2015-05-03 发布日期:2020-07-03
作者简介:杨恢先(1963-),男,湖南益阳,硕士,教授,研究方向为嵌入式系统设计和图像处理等。
基金资助:
湖南省自然科学基金(14JJ0077);湖南省教育厅一般项目(13C917)

Object Tracking Method Based on Superpixel and Local Sparse Representation

Yang Huixian, Liu Zhao, Liu Yang, Liu Fan, He Dilong

Physics and Optoelectronic Engineering College, Xiangtan University, Xiangtan 411105, China

Received:2015-03-10 Revised:2015-05-03 Published:2020-07-03

摘要/Abstract

摘要： 针对目标跟踪过程中的目标对象外观变化问题,提出一种鲁棒的基于超像素和局部稀疏表示的目标跟踪方法。在训练阶段,通过将训练帧图像分割得到的超像素聚类构造判别型外观模型;在第一帧图像中,计算目标模板的稀疏性直方图,建立生成型外观模型。在跟踪阶段,计算基于超像素的置信图,采样候选样本,计算候选样本的置信值;利用局部图块计算样本的稀疏性直方图与模板直方图的相似度。根据置信值和相似度,计算候选样本的运动模型和观测模型,得到候选样本的最大后验估计,确定目标跟踪结果。对2种外观模型都保持在线独立更新。仿真实验表明,目标发生外观变化时,算法能得到较准确和鲁棒的跟踪结果。

关键词: 超像素, 置信图, 稀疏表示, 外观模型

Abstract: Due to the appearance changing of target object in object tracking, a tracking algorithm was proposed based on superpixel and local sparse representation (SPS). In training process, a discriminative appearance model was constructed by clustering the segmented train images; sparsity-based histogram of target object was calculated to construct generative appearance model. In tracking, superpixel-based confidence map was obtained, and the confidence values of candidates was sampled and calculated; the similarity between sparsity-based histogram of candidates and target template was computed by using local patches. Then motion model and observation model of candidates according to the confidence values and similarity of candidates were computed, which obtained maximum a posterior estimate of the samples and determined the track result. Furthermore, online updating of the two appearance model was kept independently. The experimental results and evaluations demonstrate that application of SPS algorithm can obtain accurate and robust track result with the appearance variation of target object.

Key words: superpixel, confidence map, sparse representation, appearance mode

中图分类号:

TP391.9

杨恢先, 刘昭, 刘阳, 刘凡, 贺迪龙. 基于超像素和局部稀疏表示的目标跟踪方法[J]. 系统仿真学报, 2016, 28(5): 1017-1030.

Yang Huixian, Liu Zhao, Liu Yang, Liu Fan, He Dilong. Object Tracking Method Based on Superpixel and Local Sparse Representation[J]. Journal of System Simulation, 2016, 28(5): 1017-1030.

参考文献

[1] Fan Yang, Huchuan Lu, Ming-Hsuan Yang.Robust Superpixel Tracking[J]. IEEE Transactions on Image Processing (S1057-7149), 2014, 23(4): 1639-1651.
[2] Xu Jia, Huchuan Lu, Ming-Hsuan Yang.Visual Tracking via Adaptive Structural Local Appearance Model[C]// Computer Vision and Pattern Recognition (S1063-6919), Providence, Rhode Island, USA: IEEE, 2012: 1822-1829.
[3] Qing Wang, Feng Chen, Wenli Xu, et al.Object Tracking via Partial Least Squares Analysis[J]. IEEE Transactions on Image Processing (S1057-7149), 2012, 21(10): 4454-4465.
[4] B Liu, J Huang, L Yang, et al.Robust Tracking Using Local Sparse Appearance Model and k-selection[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence (S0162-8828), 2013, 35(12): 2968-2981.
[5] B Liu, L Yang, J Huang, et al.Robustand fast collaborative Tracking with two stage Sparse Optimization[C]// European Conference on Computer Vision, Heraklion (S0302-9743), 2010, Proceedings, Part IV, 2010, Crete, Greece. USA: IEEE, 2010: 624-637.
[6] H Grabner, C Leistner, H Bischof.Semi-supervised online Boosting for robust Tracking[C]// European Conference on Computer Vision, Marseille (S0302-9743), Proceedings, Part I, 2008, France. USA: IEEE, 2008: 234-247.
[7] Z Kalal, J Matas, K Mikolajczyk.P N learning: Bootstrapping binary Classifiers by structural Constraints[C]// Computer Vision and Pattern Recognition (S1063-6919), 2010, San Francisco, CA, USA. USA: IEEE, 2010: 49-56.
[8] J Kwon, K M Lee.Visual Tracking Decomposition[C]// Computer Vision and Pattern Recognition (S1063-6919), San Francisco, CA, USA. USA: IEEE, 2010: 1269-1276.
[9] X Ren, J Malik.Tracking as Repeated Figure/Ground Segmentation[C]// Computer Vision and Pattern Recognition (S1063-6919), Minneapolis, M N, USA. USA: IEEE, 2007: 1-8.
[10] X Mei, H Ling.Robust Visual Tracking using L1 minimization[C]// International Conference on Computer Vision (S1550-5499), 2009, Kyoto, Japan. USA: IEEE, 2009: 1436-1443.
[11] W Zhong, H Lu, M H Yang.Robust Object Tracking via Sparsity-based Collaborative Model[C]// Computer Vision and Pattern Recognition (S1063-6919), Rhode Island, USA. USA: IEEE, 2012: 1838-1845.
[12] Z Kalal, J Matas, K Mikolajczyk.P-N learning: Bootstrapping binary Classifiers by structural constraints[C]// Computer Vision and Pattern Recognition (S1063-6919), San Francisco, CA, USA. USA: IEEE, 2010: 49-56.
[13] F Yang, H Lu, Y W Chen.Bag of Features Tracking[C]// International Conference on Pattern Recognition (S1051-4651), Istanbul, Turkey. USA: IEEE, 2010: 153-156.
[14] David A, Ross Jongwoo, Lim Ruei-Sung Lin, et al. Incremental Learning for Visual Tracking[J]. Inter-national Journal of Computer Vision (S0920-5691), 2008, 77(1/3): 125-141.
[15] Siddiqi. TurboPixels: Fast Superpixels Using Geometric Flows[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence (S0162-8828), 2009, 31(12): 2290-2297.
[16] Comaniciu D, Meer P.Mean shift: A robust Approach toward Feature Space Analysis[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence (S0162-8828), 2002, 24(5): 603-619.
[17] 王占锋, 吴耀华, 赵林城. 删失回归模型中一个LASSO型变量选择和估计方法[J]. 应用概率统计, 2010, 26(1): 66-80.
[18] S Hare, A Saffari, P H S Torr. Struck: Structured Output Tracking with Kernels[C]// International Conference on Computer Vision (S1550-5499), 2011, Barcelona, Spain. USA: IEEE, 2011: 263-270.
[19] Y Wu, J Lim, M-H Yang.Online object tracking: A benchmark[C]// Computer Vision and Pattern Recognition (S1063-6919), 2013, Portland, OR, USA. USA: IEEE, 2013: 2411-2418.