Data-and Model-driven Attention Mechanism for Autonomous Visual Landmark Acquisition