Because of the mutual coupling of quadrotor dynamics and the sensitivity of the target image to the vehicle attitude, the issue of tracking free-motion targets in 3-dimensional space using visual servo remains challenging. For this issue, a visual servo controller based on optimized virtual image plane and image moment features is developed signed and validated. The controller enables effective tracking of arbitrarily moving targets in 3D space. The quadrotor model is decoupled by using a virtual image plane approach, which avoids image errors due to quadrotor rotation. Compared to other works using feature points, the controller uses image moment features to reduce the impact of the target’s rotation error on the velocity controller. In addition, the virtual image plane is applied to the yaw direction to construct the yaw controller, which ensures the stability of the overall visual servo system. Simulation experiments are carried out in MATLAB and Airsim to verify the effectiveness of the method.