In this paper a task-oriented approach for object tracking in large distributed camera networks is presented. This work includes three main contributions. First a generic process framework is presented, which has been designed for task-oriented video processing. Second, system components of the task-oriented framework needed for the task of multicamera person tracking are introduced in detail. Third, for an efficient task-oriented processing in large camera networks the capability of dynamic sensor scheduling by the multi-camera tracking processes is indispensable. For this purpose an efficient sensor selection approach is proposed.