|Table of Contents|

Visual Tracking Algorithm Based on a Novel Similarity Function(PDF)

南京师范大学学报（工程技术版）[ISSN:1006-6977/CN:61-1281/TN]

Issue:
2008年04期
Page:
68-72
Research Field:
Publishing date:

Info

Title:
Visual Tracking Algorithm Based on a Novel Similarity Function
Author(s):
Xia Yu; Wu Xiaojun
School of Information Engineering,Jiangnan University,Wuxi 214122,China
Keywords:
partic le filte r;  h istog ram;  sim ilar ity m easure;  H SV;  Bhattacharyya coe ffic ient
PACS:
TP391.41
DOI:
-
Abstract:
Temp la tem atch ing is an im portant link of v isua l track ing. A nove l tem plate statistical featurem atching sim -i larity criter ion is proposed w ith a track ing fram ework using particle filter. W e ighted factor is introduced to effec tive ly reduce the influence o f border no ise and background feature, and a great em inence is g iven to the im portance o f targe t feature. S im ilarity b ias m ay be got in the two com plete ly different temp la tes because of the intersection of statistica l features. The proposed m ethod can correct the bias and improv e the peak moda lity ofm a tch ing function by fusing sim ila rity based on H SV co lor system, so it can obta in the g lobal optim al so lu tion and robust tracking. Exper im enta l results show tha t tem plate m atch ing has an exce llen t peak d istr ibution, and the proposed track ing a lgor ithm exh ib its good precision and robustness in the presence o f no ise, defo rma tion and occ lusion.

References:

[ 1] 王亮, 胡卫明, 谭铁牛. 人运动的视觉分析综述[ J]. 计算机学报, 2002( 3): 225-237.
W ang Liang, H uW e im ing, Tan T ien iu. A survey o f v isual ana lysis of hum an mo tion [ J] . Ch inese Journa l o f Computers, 2002( 3): 225-237. ( in Ch inese)
[ 2] LeeK W, Ryu SW, Lee S J, et a.l M o tion based object tracking w ith m ob ile cam era[ J] . E lectron ics Lette rs, 1998, 34( 3):256-258.
[ 3] Yang C, Duraisw am i R, Dav is L. E fficient m ean-shift track ing v ia a new sim ilar ity m easure [ C] / / IEEE Compu ter So ciety
Con ference on ComputerV ision and Pattern Recogn ition. Los A lam ito s: IEEE, 2005, 1: 176-183.
[ 4] Barron J, Fleet D, Beauchem in S. Pe rfo rm ance of optica l flow techn iques [ J] . Internationa l Journal of Com puter V is ion,1994, 12( 1): 42-77.
[ 5] Com aniciu D, Ram esh V, M eer P. Kerne-l based ob ject track ing[ J] . IEEE T ransactions on Pattern Ana lysis andM achine Itelligence, 2003, 25( 5): 564-577.
[ 6] Cheung G, Baker S, Kanade T. Shape- from-silhouette o f ar ticu lated ob jects and its use for hum an body k inem atics estim ation
and m otion cap ture[ C ] / / Proceed ings of the IEEE Conference on Compu terV ision and Pa ttern Recogn ition. M ad ison: IEEE,2003, 6( 1) : 77-84.
[ 7] Nguy en H T, Sm eu lders A W M. Robust track ing us ing fo reground-backg round texture disc rim ination[ J]. In t J Comput Vision, 2006, 69( 3) : 277-293.
[ 8] Com anic iuk D, M ee r P. M ean sh ift: a robust approach tow ard feature space ana lysis[ J]. IEEE Trans Actions Pattern Anal
M ach Intel,l 2002, 24( 5): 603-619.
[ 9] Go rdon N, Salmond D. Nove l approach to non- linea r and non-Gauss ian Bayesian state estim a tion[ J]. Proceed ings of Institute Electric Eng ineer ing, 1993, 140( 2): 107-113.
[ 10] 章毓晋. 图像工程[M ]. 2版. 北京: 清华大学出版社, 2007: 168-171.
Zhang Yujin. Im ag e Eng ineer ing[M ]. 2nd ed. Be ijing: Ts inghua University Press, 2007: 168-171. ( in Chinese)

Memo

Memo:
-
Last Update: 2013-04-24