【摘要】 针对已有步态识别方法注重对步态整体特征建模，却忽略了包含细节信息的局部特征问题，提出了一种整体与局部特征相结合的双分支步态集合特征学习网络。该方法以步态轮廓集合作为输入，通过双分支网络提取两种空间尺度的步态特征，利用特征映射模块将两种特征融合得到步态表示。通过在步态数据集CASIA-B中进行了跨视角识别实验，结果表明该方法在正常行走、背包和穿大衣条件下相比主流方法 Rank-1准确率取得了有效提升。更多还原

【Abstract】 Aiming at the existing gait recognition methods that focus on modeling the global features of the gait,but ignore the local features containing detailed information,a two-branch gait set feature learning network combining global and local features is proposed.The method takes a gait silhouette set as input,extracts the gait features of two spatial scales through a two-branch network,and finally uses the feature mapping module to fuse the two features to obtain the gait representation.A series of cross-view recognition experiments are conducted on the CASIA-B gait dataset.The results show that the Rank-1 accuracy of the proposed method is effectively improved compared with the main stream methods under normal walking,carrying bags and wearing overcoats conditions.更多还原