A Mean Shift Tracking Algorithm Based on Texture Model
NING JiFeng1,2, WU ChengKe1
1.State Key Laboratory of Integrated Service Networks, Xidian University, Xi’an 710071 2.College of Information Engineering, Northwest SciTech University of Agriculture and Forestry, Yangling 712100
Abstract In Mean Shift tracking algorithm, model representation of the tracked target has great influence on the results of tracking. By analyzing image features of nine uniform texture models of Local Binary Pattern (LBP), five uniform modes related to the edge and the corner are used to represent the object model, which is called FLBP8,1, and FLBP8,1 is put into Mean Shift algorithm for object tracking. The edge is combined with its texture effectively by FLBP8,1. The key pixels of object mode are extracted automatically, and a few of them are used to represent the accurate object. Thus low computational complexity is obtained. Experimental results show that the proposed algorithm gets better performance in accuracy and speed of tracking than RGB based method under complex conditions.
[1] Comaniciu D, Ramesh V, Meer P. KernelBased Object Tracking. IEEE Trans on Pattern Analysis and Machine Intelligence, 2003, 25(5): 564575 [2] Comaniciu D, Meer P. Mean Shift: A Robust Approach toward Feature Space Analysis. IEEE Trans on Pattern Analysis and Machine Intelligence, 2002, 24(5): 603619 [3] Zhu Zhiwei, Ji Qiang. Robust RealTime Eye Detection and Tracking under Variable Lighting Conditions and Various Face Orientations. Computer Vision and Image Understanding, 2005, 98(1): 124154 [4] Charay L, Mohamed A. Tracking Multiple People with Recovery from Partial and Total Occlusion. Pattern Recognition, 2005, 38(7), 10591070 [5] Haritaoglu I, Flickner M. Detection and Tracking of Shopping Groups in Stores // Proc of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition. Kauai, USA, 2001, Ⅰ: 431438 [6] Ojala T, Valkealahti K, Oja E, et al. Texture Discrimination with MultiDimensional Distributions of Signed Gray Level Differences. Pattern Recognition, 2001, 34(3): 727739 [7] Ojala T, Pietikainen M, Maenpaa T. Multiresolution GrayScale and Rotation Invariant Texture Classification with Local Binary Patterns. IEEE Trans on Pattern Analysis and Machine Intelligence, 2002, 24(7): 971987 [8] Heikkila M, Pietikainen M. A TextureBased Method for Modeling the Background and Detecting Moving Objects. IEEE Trans on Pattern Analysis and Machine Intelligence, 2006, 28(4): 657662 [9] Chen Yachun, Chen Shuyuan. Image Classification Using Color, Texture and Regions. Image and Vision Computing, 2003, 21(9): 759776 [10] Zhang Hongming, Gao Wen, Chen Xilin, et al. Object Detection Using Spatial Histogram Features. Image and Vision Computing, 2006, 24(4): 327341 [11] Jia Jingping, Zhang Yanning, Chai Yanmei, et al. Making Tracking of Object in Image Sequences Steady Despite Complex Movement of Objects. Journal of Northwestern Polytechnical University, 2005, 23(5): 618622 (in Chinese) (贾静平,张艳宁,柴艳妹,等.目标多自由度Mean Shift序列图像跟踪算法.西北工业大学学报, 2005, 23(5): 618622) [12] Peng Ningsong, Yang Jie, Liu Zhi, et al. Automatic Selection of KernelBandwidth for MeanShift Object Tracking. Journal of Software, 2005, 16(9): 15421550 (in Chinese) (彭宁嵩,杨 杰,刘 志,等.MeanShift跟踪算法中核函数窗宽的自动选取.软件学报, 2005, 16(9): 15421550)
2007, Vol. 20 
