视频序列中人体动作识别方法的研究
首发时间:2019-08-08
摘要:针对传统监控不能实时分析场景中人的动作,只能作为事后调查的缺点,本文提出了一种对视频中人体动作识别的方法,可以实时对视频中人的动作分析理解并反馈。本文主要包括三个工作步骤:首先是使用结合空间颜色信息的背景差法检测视频中人体目标;其次选取HOG特征和多种单一特征融合的特征作为人体动作特征;最后利用SVM进行动作分类,达到对人体动作的识别。实验结果证明,该方法可以有效地对视频中人体的动作识别,具有较强的鲁棒性。
For information in English, please click here
Research on human motion recognition method in video sequence
Abstract:In view of the shortcomings of traditional monitoring, which cannot analyze human actions in the scene in real time and can only be used as an after-event investigation, this paper proposes a method of recognizing human actions in video, which can analyze and understand human actions in video in real time and give feedback.This paper mainly includes three working steps: firstly, the background difference method combined with spatial color information is used to detect human targets in video;Secondly, HOG feature and multiple single feature fusion features are selected as human motion features.Finally, SVM is used for motion classification to recognize human motion.Experimental results show that this method can effectively recognize human motion in video, and has strong robustness.
Keywords: Target detection Feature extraction The SVM classification Motion recognition
基金:
引用
No.****
动态公开评议
共计0人参与
勘误表
视频序列中人体动作识别方法的研究
评论
全部评论0/1000