摘要
本文提出了一种基于K-SVD和PCA运动属性字典学习的行为识别方法。首先采用混合高斯背景建模提取运动目标前景,并对其提取四通道光流特征;然后在字典学习阶段对于输入特征采用K-SVD学习过完备初始字典,对初始字典进行PCA变换进一步学习更为紧致的字典,减少原子间的相关性;最后采用OMP算法求解稀疏系数来表征目标的行为。利用本文提出的方法在CASIA数据集上进行测试,实验结果表明该方法具有较高的识别率。
In this paper, a behavior recognition method based on K-SVD algorithm and PCA transform attributes dictionary learning is proposed. Firstly, Guassian mixture model algorithm is adopted to extract motion object, and then extract the four-channel optical flow features of the motion object. In dictionary learning, K-SVD algorithm is used to learn an over-complete initial dictionary, and then PCA transform is used to learn a more compact and discriminative dictionary in which the atoms are uncorrelated. At last, OMP algorithm is used to compute the sparse coefficient to represent the motion of object. This method is tested in the CASIA dataset, and the result shows that our method have a higher recognition accuracy rate.
出处
《南阳理工学院学报》
2012年第6期14-17,共4页
Journal of Nanyang Institute of Technology