Fusion of color and range sensors for occupant recognition and tracking

This paper addresses the design and implementation of multisensor systems techniques that provide occupant knowledge and enable effective use of the smart space. The integration of feature selection and Bayesian classification provides consistent detection of occupants and occupant locations. Additionally, occupancy detection and activity recognition are extended through the multisensor fusion of time-of-flight (ToF) and color sensors. The existing color sensors do not provide high-resolution detection, and ToF range information and light source geometry are needed to extend this information. In this multisensor approach, predictive filter tracking techniques in the color space of the sensor response are explored in order to provide more consistent and robust detection and monitoring.