Fusion of multiple channel features for person re-identification

Abstract

Person re-identification plays an important role for automatic search of a person's presence in a surveillance video, and feature representation is a critical and fundamental problem for person re-identification. Besides, an reliable feature representation should effectively adapt to the changes of illumination, pose, viewpoint, etc. In this paper, we propose an effective feature representation called fusion of multiple channel features (FMCF) which captures different low-level features from multiple channels of HSV color space, considering the characteristics of different color channels and fusing color, texture and correlation of spatial structure. Furthermore, it takes advantage of an overlapping strategy to eliminate contrast of local cells in an image. In addition, we apply the simple weight distance metric to measure the similarity of different images, rather than metric learning which relies on a specific feature and requires more computing resources. Finally, we apply the proposed method of FMCF on the i-LIDS Multiple-Camera Tracking Scenario(MCTS) and CUHK-01person re-identification datasets, and the experimental results demonstrate that it is more robust to the variation of visual appearance.