一种多视角全局特征融合的行人再识别方法

Pedestrian re-identification is a method for identifying pedestrians across time and space through pedestrian image data. It is one of the key technologies used in construction of smart cities and smart monitoring and has attracted more and more attention from researchers in recent years. This paper proposes a method that uses the multi-convolution perspective to mine the pedestrian\'s global features in different groups, and the fuses the mined features for pedestrian re-identification. Training and verfying experiements on the open source dataset achieved an mAP of 82.1 and an accuracy rate of 92.0%, which proved the effectiveness of the proposed method.