In this paper, we investigate cooperative environmental monitoring for Pan-Tilt-Zoom(PTZ) visual sensor networks based on game theoretic cooperative control. In particular, we focus on one of the key goals of the monitoring task, i.e. monitoring environmental changes from a normal state. For this purpose, this paper first presents a novel formulation of the optimal environmental monitoring problem reflecting the above objective and characteristics of vision sensors. Then, the optimization problem is reduced to a potential game with potential function equal to the formulated objective function through an existing utility design technique, where the designed utility is shown to be computable through local computation and communication. We finally present a payoff-based learning algorithm, which refines  so that the sensors eventually take the potential function maximizes with high probability and local action constraints are dealt with. Finally, we run experiments on a testbed in order to demonstrate the effectiveness of the presented approach.