Part appearance model based on support vector machine and fuzzy k-means algorithm

doi:10.11772/j.issn.1001-9081.2015.07.2043

Abstract

Abstract:

The existing part appearance models based on Histogram of Oriented Gradient (HOG) have two defects: 1) the same cell size was used for different parts; 2) the linear Support Vector Machine (SVM) classifier can not represent the similarity of the position state and appearance model accurately. For overcoming these two defects, a part appearance model based on SVM and fuzzy k-means algorithm was built. The appearance model was composed of two classifiers: the linear SVM classifier was used to determine whether a position state belonged to human part; the similarity classifier, which was built according to the normalized Euclidean distance between the position state and the clustering center determined by fuzzy k-means algorithm, was used to calculate the similarity of the position state and appearance model. The experimental results show that the proposed appearance model can represent the appearance feature of real human part more accurately than the part appearance model built by SVM algorithm and HOG of the same cell size, and can get higher estimation accuracy when it is used to human pose estimation based on the tree-like pictorial structure model.

Key words: part appearance model, Histogram of Oriented Gradient (HOG), Support Vector Machine (SVM), fuzzy k-means algorithm, human pose estimation

摘要：

利用梯度方向直方图(HOG)建立的部位外观模型存在两个缺陷:不同部位采用相同的细胞单元尺寸,以及线性支持向量机(SVM)分类器不能准确表征部位定位状态与外观模型的相似度。为克服这两个缺陷,建立了一种基于SVM和模糊k-均值算法的部位外观模型。部位外观模型由两个分类器构成,线性SVM分类器用于判断部位定位状态是否属于人体部位,相似度分类器由部位定位状态与利用模糊k-均值算法确定的部位聚类中心的归一化欧氏距离来构造,用于计算部位定位状态与外观模型的相似度。仿真实验结果表明,与利用SVM算法和相同细胞单元尺寸建立的基于HOG特征的部位外观模型相比,新模型建立的部位外观模型能更准确地描述真实人体部位的外观特征,用于基于树形图结构模型的人体姿态估计时准确度也更高。

关键词: 部位外观模型, 梯度方向直方图, 支持向量机, 模糊k-均值算法, 人体姿态估计

CLC Number:

TP391.413

HAN Guijin. Part appearance model based on support vector machine and fuzzy k-means algorithm[J]. Journal of Computer Applications, 2015, 35(7): 2043-2046.

韩贵金. 基于支持向量机与模糊k-均值算法的部位外观模型[J]. 计算机应用, 2015, 35(7): 2043-2046.

References

[1] YANG W, WANG Y, MORI G. Recognizing human actions from still images with latent poses [C]// Proceedings of the 2010 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2010: 2030-2037.
[2] FELZENSZWALB P F, HUTTENLOCHER D P. Pictorial structures for object recognition [J]. International Journal of Computer Vision, 2005, 61(1): 55-79.
[3] THOMAS B M, HILTON A, KRUGER V, et al.Visual analysis of humans [M]. Berlin: Springer, 2011: 131-138.
[4] SAMUEL A J. Articulated human pose estimation in natural images[D]. Leeds: University of Leeds, 2012: 11-18.
[5] DALAL N, TRIGGS B. Histograms of oriented gradients for human detection [C]// Proceedings of the 2005 IEEE Conference on Com-puter Vision and Pattern Recognition. Piscataway: IEEE, 2005,1:886-893.
[6] YANG Y, RAMANAN D. Articulated human detection with flexible mixtures of parts [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2013, 35(12): 2878-2890.
[7] WANG Y, TRAN D, LIAO Z. Learning hierarchical poselets for human parsing [C]// Proceedings of the 2011 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2011: 1705-1712.
[8] JOHNSON S, EVERINGHAM M. Combining discriminative appearance and segmentation cues for articulated human pose estimation [C]// Proceedings of the 12th International Conference on Computer Vision Workshops. Piscataway: IEEE, 2009: 405-412.
[9] SAPP B, ALEXANDER T, TASKAR B. Cascaded models for articulated pose estimation [C]// Proceedings of the 11th European Conference on Computer Vision. Berlin: Springer, 2010: 406-420.
[10] CORTES C, VAPNIK V. Support-vector networks [J]. Machine Learning, 1995, 20(3): 273-297.
[11] DUNN J C. A fuzzy relative of the ISODATA process and its use in detecting compact well-separated clusters [J]. Journal of Cybernetics,1973, 3(3): 32-57.
[12] BEZDEK J C. Pattern recognition with fuzzy objective function algorithms [M]. Berlin: Springer, 1981: 65-85.
[13] ANDRILUKA M, ROTH S, SCHIELE B. Pictorial structures revisited: people detection and articulated pose estimation [C]// Proceedings of the 2009 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2009: 1014-1021.
[14] HAN G, ZHAO Y. Human pose estimation based on tree-like picture model [J]. Journal of Xi'an University of Posts and Telecommunications, 2013, 18(3): 83-86.(韩贵金,赵勇.基于树形图结构模型的人体姿态估计[J].西安邮电大学学报,2013,18 (3):83-86.)
[15] HAN G, ZHU H. Human pose estimation algorithm based on pictorial structure model [J]. Computer Engineering and Applications, 2013, 49(14): 30-33. (韩贵金,朱虹.一种基于图结构模型的人体姿态估计算法[J].计算机工程与应用,2013,49(14):30-33.)
[16] SAPP B. Pose estimation results: Buffy v2.1 [EB/OL]. [2014-12-02]. http://www.seas.upenn.edu/~bensapp/buffy_all_tiled_cps.jpg.