|本期目录/Table of Contents|

[1]田 健,王开军*,郭躬德,等.融合速度特征的压缩感知目标跟踪算法[J].南京大学学报(自然科学),2016,52(1):149-158.[doi:10.13232/j.cnki.jnju.2016.01.017]
 Tian Jian,Wang Kaijun*,Guo Gongde,et al.Object racking algorithm via speed feature in compressive tracking[J].Journal of Nanjing University(Natural Sciences),2016,52(1):149-158.[doi:10.13232/j.cnki.jnju.2016.01.017]
点击复制

融合速度特征的压缩感知目标跟踪算法()
     

《南京大学学报(自然科学)》[ISSN:0469-5097/CN:32-1169/N]

卷:
52
期数:
2016年第1期
页码:
149-158
栏目:
出版日期:
2016-02-01

文章信息/Info

Title:
Object racking algorithm via speed feature in compressive tracking
作者:
田 健王开军*郭躬德陈黎飞
(福建师范大学数学与计算机科学学院,福建省网络安全与密码技术重点实验室,福州,350007)
Author(s):
Tian Jian Wang Kaijun* Guo Gongde Chen Lifei
(School of Mathematics and Computer Science, The key laboratory of Network Security and Cryptographic Technology, Fujian Normal University, Fuzhou, 350007, China)
关键词:
目标跟踪压缩感知位移速度自适应搜索范围
Keywords:
object tracking compressive sensing displacement speed adaptive search range
分类号:
TP181
DOI:
10.13232/j.cnki.jnju.2016.01.017
文献标志码:
A
摘要:
现有快速压缩感知目标跟踪算法采用固定尺寸的搜索框搜索目标当目标快速移动时容易超出的搜索范围,导致跟踪失败。为解决此问题,提出加入目标位移速度特征的快速压缩感知跟踪算法使得搜索目标的范围自适应变化。新方法思路是利用目标在帧间的表示出目标的速度,然后将当前帧内的目标速度与前几帧的平均速度相比较,目标位移速度自适应改变搜索范围,即当目标运动速度保持稳定则保持搜索框尺寸,目标运动速度加快则增大搜索框尺寸,目标运动速度变慢则缩小搜索框尺寸,以适应目标移动速度的变化。在目标快速移动的视频集上的实验结果显示,新方法自适应地改变搜索范围,一直都能跟踪到目标,特别是当现有的压缩感知跟踪算法丢失目标时,新方法仍能比较好地跟踪到目标。
Abstract:
Object tracking has been one of the most important and active research areas in computer vision. A large number of tracking algorithms have been proposed in recent years with demonstrated success. Recently, some researchers employed compressive sensing which is originally used for efficiently acquiring and reconstructing a signal by finding solutions to underdetermined linear systems for object tracking to model object appearance. Since those existing fast compressive tracking algorithms were usually based on one fixed size search range to search for targets within a frame, it easily lead to fail when used to track one fast-moving target, especially which exceed the scope of the search. In order to solve this problem and search target successfully, a tracking algorithm based on compression sensing is proposed to use displacement speed feature to change the search range adaptively. The new method takes inter-frame motion in to account and utilizes it to represent the movement speed of target in one frame. Then the ratio between the movement speed of target in current frame and the average speed of a few frames before can be obtained to describe the scope changes of the search range. Last, the product of the ratio and the originally fixed size search range is defined as latest one in the next frame. In other word, if the target movement speed is stable, the size of search range remain; quickening up the target movement speed to increase search range or slow down to narrow, which make search range adapt to velocity. In the end of the paper, the performance of the proposed algorithm is test on datasets with the fast-motion attributes. As evaluated on challenging sequences in which the target object undergoes fast movements, the new proposed tracking algorithm performs favorably to adaptively change search range and always successfully to track target, especially when existing fast compressive tracking algorithm loses target.

参考文献/References:

[1] Yilmaz A, Javed O, Shah M, et al. Object tracking: A survey. ACM Computing Surveys, 2006, 38(4): 13.
[2] Cannons K. A review of visual tracking. Technical Report CSE-2008-07, York University, Canada, 2008.
[3] Zhang T, Ghanem B, Liu S, et al. Robust visual tracking via multi-task sparse learning. In: IEEE Conference on Computer Vision and Pattern Recognition (CVPR 2012). IEEE, 2012: 2042–2049.
[4] Bao C, Wu Y, Ling H, et al. Real time robust l1 tracker using accelerated proximal gradient approach. In: IEEE Conference on Computer Vision and Pattern Recognition (CVPR 2012). IEEE, 2012: 1830–1837.
[5] Kalal Z, Mikolajczyk K, Matas J, et al. Tracking-learning-detection. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2012, 34(7): 1409-1422.
[6] Zhang L, Dibekliogbrave H, Maaten L, et al. Speeding up tracking by ignoring features. In: IEEE Conference on Computer Vision and Pattern Recognition (CVPR 2014). IEEE, 2014: 1266-1273.
[7] Zhang K, Zhang L, Yang MH, et al. Fast compressive tracking. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2014, 36(10): 2002-2015.
[8] Zhang K, Zhang L, Liu Q, et al. Real-time compressive tracking. In: Proceedings of the 12th European Conference on Computer Vision (ECCV 2012). Spring, 2012: 864–877.
[9] Candes E J, Tao T. Decoding by linear programming. IEEE Transactions on Information Theory, 2005, 51(12): 4203-4215.
[10] 石武祯, 宁纪锋, 颜永丰等. 压缩感知跟踪中的特征特征选择与目标模型更新. 中国图象图形学报, 2014, 19(6): 932-939.
[11] 朱秋平, 颜 佳, 张 虎等. 基于压缩感知的多特征实时跟踪. 光学精密工程, 2013, 21(2): 437-444.
[12] 罗会兰, 钟宝康, 孔繁胜. 结合目标预测位置的压缩跟踪. 中国图象图形学报, 2014, 19(6): 875-885.
[13] 王松林, 项欣光. 基于压缩感知的多特征加权目标跟踪算法.计算机应用研究, 2014, 31(3): 929-932.
[14] He L, Bo Y, Zhao G, et al. Multifeatures based compressive sensing tracking. Mathematical Problems in Engineering, 2014.
[15] Yan Q, Li L. Kernel sparse tracking with compressive sensing. Computer Vision, IET, 2014, 8(4): 305-315.
[16] Wu Y, Jia N, Sun J, et al. Real-time multi-scale tracking based on compressive sensing. The Visual Computer, 2015, 31(4): 471-484.
[17] 冷亦琴, 张 莉, 杨季文等.一种基于局部稀疏线性嵌入的降维方法及其应用.南京大学学报(自然科学), 2013, 49(4): 403-410.
[18] Wu Y, Lim J, Yang M H, et al. Object tracking benchmark. IEEE Transactions on Pattern Recognition and Machine Intelligence (TPAMI), 2015.
[19] Everingham M, Gool L, Williams C, et al. The pascal visual object class (voc) challenge. International Journal of Computer Vision, 2010, 88(2): 303–338.

相似文献/References:

备注/Memo

备注/Memo:
基金项目:国家自然科学基金(61175123),福建师范大学“网络与信息安全关键理论和技术”校创新团队项目(IRTL1207)
收稿日期:2015-06-16
*通讯联系人,E-mail:wkjwang@gmail.com
更新日期/Last Update: 2016-01-27