Detecting and tracking people and their body parts in infrared

In most of today's surveillance tasks, people's actions are the focus of attention. A prerequisite for action interpretation is a stable tracking of people to build meaningful trajectories. Specifically in surveillance applications, not only trajectories on agent level are of interest, but also interpretation on the level of limbs provides important information when it comes to more sophisticated action recognition tasks. In this paper, we present an integrated approach to detect and track people and their body parts in thermal imagery. For that, we introduce a generic detection and tracking strategy that employs only local image features and thus works independently of underlying video data specifics like color information - making it applicable to both, visible and infrared data. In addition, we show how this approach serves to detect a person's body parts and extract trajectories which can be input for further interpretation purposes.