Multisensor based generation of templates for object tracking in complex scenarios

Protection and surveillance of persons and facilities is frequently associated with automatic tracking of striking objects like vehicles or persons in an image sequence. It is well known that common methods based on template matching imply problems concerning the automatic generation of realistic templates to be used in correlation based tracking systems. We propose a concept for generating templates of IR target signatures based on the fusion of IR images and classification results in laser range data. To optimize the tracking process different methods for a situation-constrained choice of the template were developed and evaluated. Results are shown by the analysis of an exemplary data set representing automatic tracking of vehicles from an airborne platform.