This paper describes our experimental study in hovering control of an unmanned helicopter with a 3D vision system instead of GPS (Global Position System). The aim of this study is to investigate methods to enable the normal operation of an autonomous helicopter when GPS does not function properly as is often experienced in cluttered environment. A 3D vision system together with a convex colored landmark is used for estimating the position of the vehicle. The estimated position is then fused with the acceleration from IMU (Inertial Measurement Unit) by Kalman filter for velocity estimation. The sensing strategy is validated by field experiments. The helicopter is stably controlled to hover within a 0.4 meter radius circle around the reference point. The hovering performance is equivalent to that when GPS is used. The good hovering performance indicates a possible application of vision system as an alternative of GPS in the case when GPS has trouble to provide reliable measurement.