Abstract

We propose a novel system for associating multi-target tracks across multiple non-overlapping cameras by an on-line learned discrim- inative appearance affinity model. Collecting reliable training samples is a ma jor challenge in on-line learning since supervised correspondence is not available at runtime. To alleviate the inevitable ambiguities in these samples, Multiple Instance Learning (MIL) is applied to learn an appearance affinity model which effectively combines three complemen- tary image descriptors and their corresponding similarity measurements. Based on the spatial-temporal information and the proposed appearance affinity model, we present an improved inter-camera track association framework to solve the “target handover” problem across cameras. Our evaluations indicate that our method have higher discrimination between different targets than previous methods.