基于3D-SIFT和SVD特征融合的视频行为识别  

Human Action Recognition Based on Multi-feature Fusion of SVD and 3D-SIFT Features

作　　者：赵雄伟[1] 刘本永[1,2] ZHAO Xiongwei LIU Benyong(College of Big Data and Information Engineering, Guizhou University, Guiyang 550025, China Institute of Intelligent hfformafion Processing, Guizhou University, Guiyang 550025, China)

机构地区：[1]贵州大学大数据与信息工程学院,贵州贵阳550025 [2]贵州大学智能信息处理研究所,贵州贵阳550025

出　　处：《贵州大学学报（自然科学版）》2017年第3期71-76,共6页Journal of Guizhou University:Natural Sciences

基　　金：国家自然科学基金资助项目(60802003);科技部国际合作项目(2009DFR10530);贵州省工业科技攻关项目(黔科合GY字(2010)3054号)

摘　　要：行为识别是视频分析的一个核心任务,而行为特征的提取与选择直接影响识别效果。针对单一特征往往受到人体外观、环境、摄像机设置等因素影响而识别效果不佳的问题,本文提出一种基于三维尺度不变特征变换与奇异值分解提取特征,通过词袋模型进行统计形成视觉词库并形成融合特征,再利用支持向量机完成特征分类的行为识别方法。实验表明三维尺度不变特征变换与奇异值分解特征融合后能很好地描述视频序列,比单独用上述两种特征或者其他传统的描述子更高效,同时也能更好地适应光照等外部影响,得到更好的识别率。Extraction and choice of motion features affects the results of the human action recognition method directly. Many factors often influence a single feature differently,such as appearance of human body,environment,and video camera. Multi-feature Fusion with Singular value decomposition（ SVD） and 3D Scale-Invariant Feature Transform Features（ 3D-SIFT） was presented in this paper wherein Bag-of words model was adopted to form fused features and SVM was used for classification. Experimental results show that fusion of 3D-SIFT and SVD can describe the video sequence well,which is more efficient than traditional single descriptors,and can better adapt to external influences such as illumination and other factors.

关 键 词：动作识别 特征融合 尺度不变特征变换 奇异值分解 词袋模型 支持向量机 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]