Vision based observation of a moving target is one of the important problems and hot issues of mobile robot system. Dynamical vision sensors which is constituted with multiple robots to obtain improved observational results has been shown to be a good substitution of single vision sensors. Thus, a new active cooperative observation (ACO) method based on two dynamical vision sensors is proposed. The most characteristics of the proposed method is that data fusion and path planning algorithm are simultaneously implemented and combined with each other by the optimal observation formation so that the influence of relative positions among MVSs and target on the cooperative observation result can be fully considered to improve the observation result. Finally, for verifying the validity and feasibility of the proposed method, experiments on a multiple rotor flying robots test-bed are conducted and analyzed, respectively.