The Study of Target Tracking Algorithm Based on Mean Shift with Adaptive Bandwidth of Anisotropic Kernel Function

Abstract

Abstract: The traditional mean shift algorithm using fixed kernels or symmetric kernel function, which have problems of target tracking lost or failure when target tracking. In order to improve the accuracy and real-time of target tracking, A target tracking algorithm is proposed based on mean shift with adaptive bandwidth of anisotropic kernel function. Firstly, this algorithm introduced the signed distance constraint function to produce the anisotropic kernel function based on signed distance kernel function. This anisotropic kernel function to meet the function value is zero in external target, and provided accurate tracking window for target moving. Secondly, The mean shift based on anisotropic kernel function is applied to target tracking to meet the sum of weight is zero of the sample points to the center point vectors in the tracking window, and calculated the mean shift window center of anisotropic kernel function template. Thirdly, anisotropic kernel function templates adaptive update implementation by similarity threshold to limit the change of the template between two sequential pictures, so as to realize real-time precise tracking. Finally, the contrast experimental results show that our algorithm has good accuracy and high real time.

Key words: adaptive bandwidth

CLC Number:

TP391

References

[1]Carreira-Perpi?án M á.Acceleration strategies for Gaussian mean shift image segmentation[C]. Proc of IEEE Conf. on Computer Vision and Pattern Recognition, Washington, DC: IEEE Computer Society, 2006, 1: 1160-1167.
[2]Carreira-Perpinan M A, Williams C.On the Number of Modes of a Gaussian Mixture[C]. In Proc. of the 4th Int. Conf. on Scale-Space theories in Computer Vision, Heidelberg: Springer Berlin, 2003, 2695: 625- 640.
[3]Peng N S, Yang J, Liu Z.Mean shift blob tracking with kernel histogram filtering and hypothesis testing[J].Pattern Recognition Letters, 2005, 26(5):605-614
[4]Liu J M, Han M, Wang Z Z.Research on Multi-target Tracking Algorithm Based on Adaptive Background Updating and MSPF[J].Journal of Computational Information Systems, 2012, 8(24):10081-10088
[5]Zhou Xiaolong, Shen Henghua, He Bingwei.Brith intensity estimation method for mulit-target video tracking[J].Journal of Computer-Aided Design and Computer Graphics(计算机辅助设计与图形学报). 2014, 26(12): 2223-2231
[6]Mohsen A, Homayoun M N.A Robust Mean-Shift Tracking Using GMM Background Subtraction[J].Journal of Basic and Applied Scientific Research, 2013, 3(4):596-607
[7]Wang J, Thiesson B, Xu Y.Image and videosegmentation by anisot ropic kernel mean shift [C].Proc of European Conf on Computer Vision, 2004: 5742583.
[8]Zhang Jinhua, Hou Yan, Yang Jun.Image restoration based on nonlocal total variation[J].量子电子学报, 2015, 32(3):278-282
[9]Yilmaz A, Javed O, Shah M.Object tracking: A survey[J].Acm Computing Surveys(CSUR), 2006, 38(4):1-45
[10]Han Ming, Liu Jiaomin, Meng Junying.Local Energy Information Combined with Improved Signed Distance Regularization Term for Image Target Segmentation Algorithm[J]. Journal of Electronics & Information Technology(电子与信息学报), 2015, 37(9):2047-2054.
[11]Osher S, Sethian J A.Fronts propagating with curvature-dependent speed: algorithms based on Hamilton-Jacobi formulations[J].Journal of Computational Physics, 2008, 79(1):12-49
[12]Wang Xiaofeng.The Study of level set Methods and their applications in image segmentation[D].BeiJing: Dissertation of University of Science and Technology of China, 2009, 10-20.
[13]Yang Sa, Zheng Zhishuo.Medical image registration algorithm based on sparse random projection and SIFT transform[J].量子电子学报, 2015, 32(3):283-289
[14]Li Jun, Yang Xin, Shi Pengfei.A fast level set approach to image segmentation based on mumford-shah model[J]. Chinese Joutnal of Computers(计算机学报), 2002, 25(11): 1175-1183.
[15]Zhang Fengjun, Zhao Ling, An Guocheng, et al.Mean shift tracking algorithm with scale adaptation[J].计算机研究与发展, 2014, 51(1):215-224
[16]Han Ming, Liu Jiaomin, Meng Junying, et al.A Modeling and Target Detection Algorithm Based on Adaptive Adjustment K-ρ for Mixture Gaussian Background[J].电子与信息学报, 2014, 36(8):2023-2027