Target Tracking in the Presence of Intermittent Measurements by a Sparsely Distributed Network of Stationary Cameras

An observer and predictor framework is developed for tracking a moving object using a sparsely distributed network of stationary cameras. The sparsity of the camera network is representative of complex environments where continuous image feedback is not available, which occurs when the target is transitioning between fields-of-view in the camera network or the target is occluded by objects in the scene. Using a Lyapunov-based stability analysis, estimates of the object's pose and motion model are proven to converge, provided the current network configuration satisfies certain dwell-time conditions (i.e., minimum time the object needs to be observed and maximum time the object may remain outside a feedback region). This approach allows for areas of the network that may cause instabilities in the pose estimates, based on the dwell-time conditions, to be identified ahead of time and offers insight into how the network configuration could be augmented to ensure stability of the target's pose estimates.