Visual tracking of patients with specific adverse conditions such as epileptic seizures is an important task related to the prevention of unwanted medical situations and events. Previously, we have developed algorithms for contactless patient tracking based on optical flow analysis. In this work, we address some of the challenges faced by the single-camera tracking system and expand its functionalities by employing simultaneous input from multiple cameras. Methods. We propose a new approach of fusing multiple camera sensors. It uses a proprietary motion-group parameters reconstruction algorithm and includes scenarios of both overlapping and non-overlapping fields of view. In the first case, the simultaneous tracking within the overlapping field evolves from independent tracking by each camera, to a synchronized tracking by a set of cameras. This is achieved by automated reinforcement learning and simultaneously applying the interdependences between the cameras. In addition, outside the overlapping areas the algorithm can transfer tracking from one camera to another provided tree-type of topology between the areas is present. Results. We demonstrate that synchronous, multi-camera tracking scenarios provide improvements on both real-world and simulated tests. This new approach allows improving the accuracy and robustness of the original methods, to extend the tracking coverage and to provide other beneficial effects, such as more precise detection of fast-moving objects. The proposed method is compared with other algorithms used in the field.