Person re-identification using semantic color names and RankBoost

We address the problem of appearance-based person re-identification, which has been drawing an increasing amount of attention in computer vision. It is a very challenging task since the visual appearance of a person can change dramatically due to different backgrounds, camera characteristics, lighting conditions, view-points, and human poses. Among the recent studies on person re-id, color information plays a major role in terms of performance. Traditional color information like color histogram, however, still has much room to improve. We propose to apply semantic color names to describe a person image, and compute probability distribution on those basic color terms as image descriptors. To be better combined with other features, we define our appearance affinity model as linear combination of similarity measurements of corresponding local descriptors, and apply the RankBoost algorithm to find the optimal weights for the similarity measurements. We evaluate our proposed system on the highly challenging VIPeR dataset, and show improvements over the state-of-the-art methods in terms of widely used person re-id evaluation metrics.

1. Introduction

Person re-identification is a critical problem in a video surveillance system. Its goal is to re-identify a person in different locations across multiple, potentially non-overlapping, cameras. Due to the unreliable spatial and temporal information, appearance-based person re-id has been drawing an increasing amount of attention in recent computer vision research. The common assumptions for this task include: a) the finer biometric cues (e.g. face, or iris) are not available due to the low image resolution; b) The targets of interest do not change their clothes across different cameras. In other words, appearance-based person re-id relies on the information provided by the visual appearance of human body and clothing. It is a highly challenging problem since human appearance usually exhibits large variations across different cameras. This variation is due to variability in backgrounds, sensor characteristics, lighting conditions, view-points, and human poses. Besides, distinct people may look similar if they wear clothes with the same color, which in turn increases the difficulty of finding correct associations.