Several non-linear state estimation methods such as extended Kalman filter, cubature Kalman filter, and unscented Kalman filter are used to track objects in visual sensor networks. These conventional non-linear state estimation methods require the accurate knowledge of the object's initial conditions, process and measurement models, and corresponding noise characteristics. Often, the object trackers used in a visual sensor networks may not be provided with this knowledge. In this work, we propose a square root cubature H∞ information Kalman filter (SCHIF) based distributed object tracking algorithm. The H∞ method requires neither the exact knowledge of noise characteristic nor accurate process model. The information filters can be used without the knowledge of accurate initial conditions and it also makes the measurement update step computationally less complex in the distributed process. Finally, the square root version makes the filter numerically stable. Furthermore, the cameras in the network exchange their local estimates with other cameras. In the last step, the cameras fuse the received local estimates to obtain a global estimate of the object. Hence, the proposed method constitutes a more robust and efficient solution for the targeted application compared to the traditional methods.