This article addresses the problem of tracking moving ground target with unknown motion for quadrotor in GPS-denied environments. Firstly, an IBVS controller, which utilizes virtual image features to decouple the translational and rotational motion of quadrotor, is proposed, effectively addressing underactuated issues. Secondly, the research solely relies on visual sensors, which intelligently extract and utilize image information to design the controller, thereby significantly reducing the dependence on external hardware and demonstrating flexible tracking capabilities in complex scenes. In addition, prescribed performance control enables efficient operation even under unknown target motion conditions.