A multi-camera approach to tracking and localization of people with coexisting robots

People tracking and localization in situations with coexisting service robots is a challenging task due to the non-rigid nature of human body, complex background and frequent occlusions among people and robots. A multi-view approach is proposed to people-tracking and localization whilst the robotpsilas capability of perceiving environment is utilized. Initiated by person detection, tracking within single camera is performed by particle filtering based on the weighed color histogram. A multi-camera collaboration method is proposed to track multiple people and obtain integrated people location based on ground plane homographics. This approach enables people-tracking using multiple cameras with overlapped field-of-views and outperforms the single-view approaches in tracking accuracy especially in situations with partial occlusions. The effectiveness of proposed approach is validated by an implementation in a home-care surveillance system, in which people are supported by robot assistants for an easy living.