首页 > 论文 > 光学学报 > 37卷 > 8期(pp:815001--1)

基于核循环结构的自适应特征融合目标跟踪

Adaptive Feature Fusion Object Tracking Based on Circulant Structure with Kernel

  • 摘要
  • 论文信息
  • 参考文献
  • 被引情况
  • PDF全文
分享:

摘要

视频跟踪中, 使用单一特征对目标进行描述难以适应复杂场景的变化, 目标的尺度变化、形变、遮挡等因素易导致跟踪失败。为提高跟踪的稳健性, 基于核循环结构, 提出一种自适应特征融合和模型更新的跟踪方法, 并引入尺度更新机制。首先利用目标的灰度特征和局部二值模式特征分别计算滤波响应图, 依据响应图的峰值旁瓣比(PSR)自适应地分配权值并加权融合, 从而估计目标的最佳位置。然后根据融合后响应图的PSR来判断跟踪质量, 据此决定是否更新模型。最后在目标位置周围提取方向梯度直方图特征构建尺度金字塔, 训练尺度相关滤波器, 用来估计目标的最佳尺度。实验选取标准测试数据集中具有光照变化, 遮挡和尺度变化的视频序列进行实验, 结果表明, 该算法能够实现对目标的稳定跟踪, 并且在距离精度和成功率上均优于对比算法。

Abstract

In video tracking, the use of a single feature to describe the target is difficult to adapt to the changes in complex scenes. Futhermore, the scale change, deformation, occlusion of target and other factors will lead to tracking failure. In order to improve the robustness of tracking, an adaptive feature fusion and model updating tracking algorithm is proposed based on the circulant structure of with kernel, and the scale updating mechanism is also introduced. Firstly, the response maps are calculated using the gray and local binary pattern features of the target respectively and fused by the weights assigned according to the peak to sidelobe ratio(PSR), and the best location is estimated. The PSR of the fused response map is also used to judge the tracking quality to decide whether to update the model. Finally, according to the scale pyramid constructed with the histograms of oriented gradients features extracted around the target location, the scale correlation filter is trained to estimate the optimal scale of the target. The experiment selects the sequences with illumination variations, occlusion and scale changes from the visual tracker benchmark datasets. The results show that the proposed algorithm can track the target robustly in complex scenes, and the distance precision and success rate are also superior to the compared algorithms.

投稿润色
补充资料

中图分类号:TP391

DOI:10.3788/aos201737.0815001

所属栏目:机器视觉

基金项目:江苏省博士后科研资助计划(1601181B)

收稿日期:2017-02-27

修改稿日期:2017-04-11

网络出版日期:--

作者单位    点击查看

赵高鹏:南京理工大学自动化学院, 江苏 南京 210094
沈玉鹏:南京理工大学自动化学院, 江苏 南京 210094
王建宇:南京理工大学自动化学院, 江苏 南京 210094

联系人作者:赵高鹏(zhaogaopeng@njust.edu.cn)

备注:赵高鹏(1983-),男, 博士, 讲师, 主要从事计算机视觉方面的研究。

【1】Zhang Huanlong, Hu Shiqiang, Yang Guosheng. Video object tracking based on appearance models learning[J]. Journal of Computer Research and Development, 2015, 52(1): 177-190.
张焕龙, 胡士强, 杨国胜. 基于外观模型学习的视频目标跟踪方法综述[J]. 计算机研究与发展, 2015, 52(1): 177-190.

【2】Sun Rui, Huang Jingru, Ding Wenxiu. A real-time object tracking algorithm based on subspace learning[J]. Opto-Electronic Engineering, 2015, 42(2): 52-58.
孙 锐, 黄静茹, 丁文秀. 一种基于子空间学习的实时目标跟踪算法[J]. 光电工程, 2015, 42(2): 52-58.

【3】Yang F, Lu H C, Yang M H. Robust superpixel tracking[J]. IEEE Transactions on Image Processing, 2014, 23(4): 1639-1651.

【4】Liu Wenzhuo, Yuan Guanglin, Xue Mogen. Robust fast visual tracking based on two-stage sparse representation[J]. Acta Optica Sinica, 2016, 36(12): 1215001.
刘文琢, 袁广林, 薛模根. 基于两阶段稀疏表示的鲁棒快速视觉跟踪[J]. 光学学报, 2016, 36(12): 1215001.

【5】Babenko B, Yang M H, Belongie S. Robust object tracking with online multiple instance learning[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2011, 33(8): 1619-1632.

【6】Xue Yizhe, Wang Tuo. Object-tracking method based on improved cost-sensitive adaboost[J]. Journal of Image and Graphics, 2016, 21(5): 544-555.
薛一哲, 王 拓. 基于代价敏感Adaboost目标跟踪[J]. 中国图象图形学报, 2016, 21(5): 544-555.

【7】Li Y, Zhu J K. A scale adaptive kernel correlation filter tracker with feature integration[C]. Proceedings of 13th European Conference on Computer Vision, 2014, 8926: 254-265.

【8】Bolme D S, Beveridge J R, Draper B A, et al. Visual object tracking using adaptive correlation filters[C]. Proceedings of 2010 IEEE Conference on Computer Vision and Pattern Recognition, 2010: 2544-2550.

【9】Henriques J F, Caseiro R, Martins P, et al. Exploiting the circulant structure of tracking-by-detection with kernels[C]. Proceedings of 12th European Conference on Computer Vision, 2012, 7575: 702-715.

【10】Henriques J F, Caseiro R, Martins P, et al. High-speed tracking with kernelized correlation filters[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2015, 37(3): 583-596.

【11】Danelljan M, Khan F S, Felsberg M, et al. Adaptive color attributes for real-time visual tracking[C]. Proceedings of 2014 IEEE Conference on Computer Vision and Pattern Recognition, 2014: 1090-1097.

【12】Ojala T, Pietikainen M, Harwood D. Performance evaluation of texture measures with classification based on kullback discrimination of distributions[C]. Proceedings of 12th IAPR International Conference on Pattern Recognition, 1994, 1: 582-585.

【13】Rifkin R, Yeo G, Poggio T. Regularized least-squares classification[J]. Nato Science Series Sub Series III Computer and Systems Sciences, 2003, 190: 131-154.

【14】Danelljan M, Hger G, Khan F S, et al. Accurate scale estimation for robust visual tracking[C]. Proceedings of the British Machine Vision Conference 2014, BMVA Press, 2014.

【15】Hare S, Saffari A, Torr P H S. Struck:structured output tracking with kernels[C]. Proceedings of IEEE International Conference on Computer Vision, 2011: 263-270.

【16】Babenko B, Yang M H, Belongie S. Visual tracking with online multiple instance learning[C]. Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, 2009: 983-990.

【17】Kalal Z, Mikolajczyk K, Matas J. Tracking-learning-detection[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2012, 34(7): 1409-1422.

【18】Zhang K H, Zhang L, Yang M H. Real-time compressive tracking[C]. Proceedings of 12th European Conference on Computer Vision, 2012, 7574: 864-877.

【19】Wu Y, Lim J, Yang M H. Online object tracking: a benchmark[C].Proceedings of 26th IEEE Conference on Computer Vision and Pattern Recognition, 2013:2411-2418.

引用该论文

Zhao Gaopeng,Shen Yupeng,Wang Jianyu. Adaptive Feature Fusion Object Tracking Based on Circulant Structure with Kernel[J]. Acta Optica Sinica, 2017, 37(8): 0815001

赵高鹏,沈玉鹏,王建宇. 基于核循环结构的自适应特征融合目标跟踪[J]. 光学学报, 2017, 37(8): 0815001

您的浏览器不支持PDF插件,请使用最新的(Chrome/Fire Fox等)浏览器.或者您还可以点击此处下载该论文PDF