Person Re-Identification Based on Multi-Level and Multi-Feature Fusion

Due to variations in pose and illumination condition, the appearance of can be significantly different in different cameras and the performance of person re-identification is degraded. In this paper, a person re-identification based on multi-level and multi-feature fusion for this phenomenon is proposed. Firstly, we divided each sample into three parts and multi-layer sampling. Secondly, we extracted color histogram and Weber Local Descriptor (WLD) from each part and fused. Lastly, We use weighted Euclidean distance to complete person re-identification across non-overlapping cameras. Experimental results on two public datasets show that the proposed method outperforms the state-of-art approaches.