Kernel Cross-View Collaborative Representation based Classification for Person Re-Identification

TL;DR

This paper introduces Kernel Cross-View Collaborative Representation based Classification (Kernel X-CRC), a nonlinear method that improves person re-identification accuracy across different camera views by representing images in a shared high-dimensional feature space.

Contribution

It proposes a novel kernel-based collaborative representation model that effectively handles nonlinear appearance changes and the small-sample-size problem in person re-identification.

Findings

Achieves state-of-the-art rank-1 accuracy on PRID450S and GRID datasets.

Outperforms existing methods on multiple person re-identification benchmarks.

Demonstrates robustness with both high-dimensional and low-dimensional feature representations.

Abstract

Person re-identification aims at the maintenance of a global identity as a person moves among non-overlapping surveillance cameras. It is a hard task due to different illumination conditions, viewpoints and the small number of annotated individuals from each pair of cameras (small-sample-size problem). Collaborative Representation based Classification (CRC) has been employed successfully to address the small-sample-size problem in computer vision. However, the original CRC formulation is not well-suited for person re-identification since it does not consider that probe and gallery samples are from different cameras. Furthermore, it is a linear model, while appearance changes caused by different camera conditions indicate a strong nonlinear transition between cameras. To overcome such limitations, we propose the Kernel Cross-View Collaborative Representation based Classification (Kernel X-CRC) that represents probe and gallery images by balancing representativeness and similarity nonlinearly. It assumes that a probe and its corresponding gallery image are represented with similar coding vectors using individuals from the training set. Experimental results demonstrate that our assumption is true when using a high-dimensional feature vector and becomes more compelling when dealing with a low-dimensional and discriminative representation computed using a common subspace learning method. We achieve state-of-the-art for rank-1 matching rates in two person re-identification datasets (PRID450S and GRID) and the second best results on VIPeR and CUHK01 datasets.