【摘要】 针对步态能量图(GEI)和图像序列的Radon变换可以表征图像能量的特点,提出这两种形式的能量特征相融合的方法进行身份识别.在周期分割后的特征提取阶段分别使用GEI结合行列相结合的二维主成分分析((2D)2PCA)方法和对步态序列图像进行Radon变换,在周期模板构造后用列方向的二维主成分分析(2DPCA)降维方法进行数据压缩.在识别阶段,采用多视角及多特征在决策层的融合方法.应用上述方法在CASIA步态数据库上进行实验,结果表明所提的步态识别方法具有较高的识别性能.更多还原

【Abstract】 From the fusion of energy representation of both static gait energy images(GEI) and dynamic energy representation of Radon transformation of image sequences,a novel gait recognition method was proposed.Feature extraction from GEI which described the gait characteristics was obtained from periodic sequence images,while Radon transform of the gait image sequence from one cyclical gait character was portrayed by using template structure.Following these,feature reduction was effected using two directional two dimensional principal component analysis((2D)2 PCA) and column directional two dimensional principal component analysis(2DPCA) respectively.Fusion of multiple-views as well as multiple-characteristics was accomplished in the final gait recognition stage.The proposed gait recognition was evaluated method on CASIA gait database,and the experimental results demonstrate that our approach has encouraging recognition performance.更多还原