CHEN Sheng-di,HE Bing-qian,CHEN Si-yu,et al.Human Action Recognition based on Spatio-Temporal Interest Point[J].Journal of Chengdu University of Information Technology,2018,(02):143-148.[doi:10.16836/j.cnki.jcuit.2018.02.007]
基于时空兴趣点的人体动作识别
- Title:
- Human Action Recognition based on Spatio-Temporal Interest Point
- 文章编号:
- 2096-1618(2018)02-0143-06
- Keywords:
- computer applications technology; image processing and graphics; spatial and temporal interest points; action recognition; harris-laplace; 3D-SIFT; feature extraction
- 分类号:
- TP391.41
- 文献标志码:
- A
- 摘要:
- 人体动作识别在计算机视觉研究和模式识别领域中逐渐成为一个研究热点。提出一种基于Harris-Laplace时空兴趣点结合3D-SIFT描述子,通过Bag-of-feature构建词袋的方法,并应于用人体动作识别。针对传统Harris算法提取出的兴趣点冗余,所以采用Harris-Laplace算法提取时空兴趣点。3D-SIFT描述子能更好地描述视频序列的本质特征,并且比传统的描述子更有效,Bag-of-feature词袋法表征特征,采用改进的K均值(K-Means)聚类算法进行聚类,最后采用多分类支持向量机(SVM)进行一对一、一对多的分类策略并进行比较。在KTH公开运动数据集上进行实验测试,实验结果证明提出的人体动作识别方法的有效性和鲁棒性。
- Abstract:
- Human action recognition are increasingly attracting much attention from computer vision and pattern recognition researchers. This paper presents a method based on Harris-Laplace algorithm combined with 3D-SIFT descriptor, and a Bag-of-feature approach is used to represent videos. The Harris-Laplace algorithm is used to extract the spatial and temporal interest points. The3D-SIFT descriptor can better describe the essential characteristics of the video sequence and it is more effective than the traditional descriptor. The K-Means approach is used for clustering. Finally, the support vector machine(SVM)is used as the classifier for human action recognition. One-vs-one and one-vs-rest classification strategies are used and the comparison is made. The experiment on the public database KTH proves the effectiveness and robustness of this method.
参考文献/References:
[1] 张博宇,刘家锋,唐降龙.一种基于时空兴趣点的人体动作识别方法[J].自动化技术与应用,2009,28(10):75-78.
[2] 王见,陈义,邓帅.基于改进SVM分类器的动作识别方法[J].重庆大学学报,2016,39(1):12-16.
[3] Watananbe T,Tanaka T.Vein authentication using color information and image matching with hight performance on natural light[C].Proceedings of International Joint Conference.Fukuoka:Fukuoka International Congress Center,2009:3625-3629.
[4] Yang M Y,Cao Y,McDonald J. Fusion of camera images and laser scans for wide baseline 3D scene alignment in urban environments[J]. ISPRS Journal of Photogrammetry and Remote Sensing,2011,66:52-61.
[5] An Hangxing,Meng Lingjun,Zhao Lin,et al. Long-distance transmission and high-speed and real-time storage technology of image data[J]. Video Engineering,2013,37(3):175-178.
[6] Cao Y,McDonald J. Improved feature extraction and matching in urban environments based on 3D view point normalization[J]. Computer Vision and Image Understanding,2012,116:86-101.
[7] W He. Recognition of human activities using a multclass relevance vector machine[J]. Optical Engineering SPIE,2012.
[8] Anoop Cherian,Piotr Koniusz,Stephen Gould. Higher-order Pooling of CNN Features via Kernel Linearization for Action Recognition[C]. IEEE Winter Conference on Applications of Computer Vision,2017:130-137.
[9] Laptev I. On Space-Time Interest Points[J]. International Journal of Computer Vision,2005,64(2/3):432-439.
[10] Park S,Aggarwal J K. A Hierachical Bayesian Network for Even Recognition of Human Actions and Interactions[J]. Multimedia Systems,2004,10(2):164-179.
[11] 丁松涛,曲仕茹. 基于改进时空兴趣点检测的人体行为识别算法[J]. 西北工业大学学报,2016,34(5):886-892.
[12] Guan T,Wang C. Registration based on scene recognition and natural features tracking techniques for widearea augmented reality systems[J].IEEE Transaction on Multimedia,2009,11(8):1393-1406.
[13] C Schuldt,I Laptev,B Caputo. Recognizing Human Actions: A Local SVM Approach[C]. Proceedings of the 17th Interenational Conference on Pattern Recognition.ICPR,2004.
[14] Sushideep Narayana. Action Recognition form video[EB/OL]. https://github.com/Sushirdeep/Action-Recognition-from-Videos/blob/master/ProjectReport/ActiionRecogntionReport_Sushirdeep.pdf.
[15] Arthur D, Vassilvitskii S. k-means++: the advantages of careful seeding[C]. Eighteenth Acm-Siam Symposium on Discrete Algorithms. Society for Industrial and Applied Mathematics, 2007:1027-1035.
[16] KTH Database[EB/OL].http://www.nada.kth.se/cvap/actions/.
[17] 付朝霞,王黎明. 基于时空兴趣点的人体行为识别[J]. 微电子学与计算机,2013,30(8):28-35.
[18] Dufournaud Y,Schmid C,Horaud R. Matching Images with Different Resolutions[C]. The IEEE Conferences on Computer Vision and Pattern Recognition,Hilton Head Island,South Carolina,2000.
[19] Mokhtarian F,Suomela R. Curvature Scale Space Based Image Corner Detection[C]. EuropeaanSignal Processing Conference,Island of Rhodes,Greece,1998.
[20] Lindeberg T.Feature Detection with Automatic Scale Selection[J]. International Journal of Computer Vision,1998,30(2):79-116.
[21] Scovanner P, Ali S, Shah M. A 3-dimensional sift descriptor and its application to action recognition[C]. DBLP, 2007:357-360.
[22] Hartigan J A, Wong M A. A K-means clustering algorithm[J]. Applied Statistics, 1979, 28(1):100-108.
[23] 梁宇宏,张欣. 对遗传算法轮盘赌选择方式的改进[J]. Information Technology,2009,33(12):127-129.
[24] 夏桂梅,曾建潮. 一种基于轮盘赌选择遗传算法的随机微立群算法[J].计算机工程与科学,2007,29(6):51-54.
[25] 蔡军,邹鹏,沈弼龙,等.改进轮盘赌策略的反馈式模糊测试方法[J]. 四川大学学报,2016,48(2):133-137.
[26] 万士宁. 基于卷积神经网络的人脸识别研究与实现[D].成都:电子科技大学,2016.
相似文献/References:
[1]冯丽君,王 燮,王小东.基于FAHP的气象仿真服务逼真度评价研究[J].成都信息工程大学学报,2016,(02):168.
FENG Li-jun,WANG Xie,WANG Xiao-dong.Evaluation of Meteorological Simulation Service Fidelity based on Fuzzy AHP[J].Journal of Chengdu University of Information Technology,2016,(02):168.
[2]吴东华,常 征,何 嘉.基于用户行为序列模式的性别分析与预测[J].成都信息工程大学学报,2016,(增刊1):7.
备注/Memo
收稿日期:2017-07-11基金项目:四川省教育厅重点科研资助项目(17ZA0064)