欢迎您！东篱公司
退出
-
申报数据库
-
立项数据库
-
成果数据库
-
项目获奖数据库
中文摘要：
文中提出一种基于时空特征点轨迹的动作识别方法.首先为了克服局部时空特征时间信息缺失的问题,该方法采用KLT跟踪器对时空局部特征进行跟踪,将得到的时空特征跟踪轨迹作为基本的处理、描述单元.与局部时空特征相比,它能在更长的时间尺度上对运动进行描述,进而更好地捕获运动的动态变化与转变过程.其次在时空特征轨迹基础上,该方法提出了轨迹相对位置、相对速度关系元来对轨迹之间的关系进行建模.对轨迹之间的关系进行建模有助于捕获不同动作在特征分布上存在的一些比较稳定的模式.最后利用多核学习方法融合多种特征来训练动作分类器.在交互动作数据库上对提出的方法进行了实验,实验结果证明了方法的有效性.
英文摘要：
This paper proposes an approach to recognize human activities,which is based on tracking trajectories of local spatio temporal feature points.To make up for the temporal informa tion loss of local features,this paper uses the KLT feature tracker to track each spatial temporal local feature and treats the tracked feature trajectory snippets as the basic processing and describing unit.Compared with local spatio temporal feature,it can capture the motion information of an action pattern in a longer time scale and better describe the dynamic characteristics and transitions of motion.As to the relationship modeling among feature trajectory snippets,we believe that there exist some stable feature distribution patterns in an action video clip,which lie in the interconnec tion of position and velocity between local features,so we propose the relative position relation and relative velocity relation descriptors to capture this kind of relation.Experimental results on the UT Interaction dataset are provided to demonstrate the effectiveness and robustness of our approach.
同期刊论文项目
同项目期刊论文
Image classification by non-negative sparse coding, correlation constrained low-rank and sparse deco
Undoing the codebook bias by linear transformation with sparsity and F-norm constraints for image cl
期刊信息
位置：