基于局部特征点方法的动作识别研究

Research on action recognition based on the method of local features

利用局部特征点方法进行视频中的动作识别研究,通过对尺度空间理论以及多种经典的局部特征点检测方法的深入分析,将SIFT算法引入到视频研究领域,提出了一种全新而高效的视频特征点检测算法。在此基础上,设计了合理的机器学习方法对局部特征点进行训练,利用训练出的通用动作模式在视频中进行动作识别。

The aim of this paper is to address recognition of human actions using the method of local features. We generalize SIFT method to the domain of video research, and present a new local feature detection algorithm. The presented algorithm is validated to be high performance. Based on this new algorithm, we design a reasonable method of machine learning to obtain general patterns of human action, and recognize human actions in videos using the given patterns.