基于搜索算法K-Means动作关键特征序列的行为识别方法
Action Recognition Based on Key Poses Sequences with Searching-Based K-Means Algorithm

作者: 殷 鑫 :苏州大学数学科学学院,江苏 苏州; 龚声蓉 , 刘纯平 :苏州大学计算机科学与技术学院,江苏 苏州;

关键词: 行为识别集合论平均聚类算法动态时间规整算法Action Recognition Set Theory K-Means Dynamic Time Warping

摘要:
行为识别是近年来计算机视觉领域的一个研究热点。本文在当今已有的行为识别算法的基础之上进行优化改进。通过基于代数理论的背景减法提取轮廓并进行姿势表达、通过聚类算法提取动作关键特征,并基于DTW动态时间规整算法完成动作识别。由于原始K-means算法中聚类结果对于初值的依赖性,我们引入基于搜索算法的K-means聚类算法,避免了初值对聚类结果的影响。通过在国际主要数据库上的实验,达到了较高的准确率和稳定度,并能够实现在线实时识别。

Abstract: Vision-based human action recognition is currently one of the most active research fields. Action recognition is a cross-disciplinary field which contains theories of image processing, computer vi-sion and artificial intelligence. Firstly, we get contours and pose presentation through background subtraction algorithm based on algebra theory and then we get the key poses of action through improved searching-based K-means algorithm. Finally actions are recognized through dynamic time warping algorithm. Experimental results on the main datasets show suitability for online recognition and real-time scenarios.

文章引用: 殷 鑫 , 龚声蓉 , 刘纯平 (2015) 基于搜索算法K-Means动作关键特征序列的行为识别方法。 图像与信号处理, 4, 1-10. doi: 10.12677/JISP.2015.41001

参考文献

[1] 杜友田, 陈峰, 徐文立, 李永彬 (2007) 基于视觉的人的运动识别综述. 电子学报, 1, 84-90.

[2] 李瑞峰, 王亮亮, 王珂, 等 (2014) 人体动作行为识别研究综述. 模式识别与人工智能, 1, 35-48.

[3] 李莉莉 (2010) 基于视频的指尖检测与跟踪算法及实现. 内蒙古大学, 呼和浩特.

[4] 王忠礼 (2008) 智能交通系统车辆检测算法研究. 哈尔滨工业大学, 哈尔滨.

[5] 金慧 (2005) 基于多视图像的三维建模方法研究. 南京理工大学, 南京.

[6] Dedeoglu, Y., Toreyin, B., Gudukbay, U. and Cetin, A. (2006) Silhouette-based method for object classification and human action recognition in video. In: Huang, T., Sebe, N., Lew, M., Pavlovic, V., Kolsch, M., Galata, A. and Kisacanin, B., Eds., Computer Vision in Human-Computer Interaction, Springer, Berlin, Heidelberg, Volume 3979 of Lecture Notes in Computer Science, 64-77.

[7] Cheema, S., Eweiwi, A., Thurau, C. and Bauckhage, C. (2011) Action recognition by learning discriminative key poses. IEEE International Conference on Computer Vision Workshops (ICCV Workshops), Barcelona, 6-13 November 2011, 1302-1309.

[8] Baysal, S., Kurt, M. and Duygulu, P. (2010) Recognizing human actions using key poses. 20th International Conference on Pattern Recognition (ICPR), Istanbul, 23-26 August 2010, 1727-1730.

[9] 严勇 (2007) 数据挖掘中聚类分析算法研究与应用. 电子科技大学, 成都.

[10] 李邵梅, 刘力雄, 陈鸿昶, 等 (2008) 实时说话人辨识系统中改进的DTW算法. 计算机工程, 4, 218-219.

[11] Chaaraoui, A.A., Climent-Pérez, P. and Flórez-Revuelta, F. (2013) Silhouette-based human action rec-ognition using sequences of key poses. Pattern Recognition Letters, 34, 1799-1807.

[12] Blank, M., Gorelick, L., Shechtman, E., Irani, M. and Basri, R. (2005) Actions as space-time shapes. Tenth IEEE International Conference on Computer Vision, Beijing, 17-21 October 2005, Vol. 2, 1395-1402.

分享
Top