A switched systems approach to image-based localization of targets that temporarily leave the field of view

Vision sensors provide rich information about the environment and can be utilized for tracking the position of moving objects from a moving camera. This paper presents the development of dwell time conditions to guarantee convergence to an ultimate bound of state estimation errors for a class of vision based observers with intermittent measurements. Bounds are developed on the unstable growth of the estimation errors during the periods when the object being tracked is not visible. A Lyapunov analysis for the switched system is performed to develop an inequality in terms of the duration of time the observer can view the moving object and the duration of time the object is out of the field of view.