In this paper, a control system for autonomous landing of UAV at a desired position using vision system is constructed and verified through flight experiments. In the UAV, a camera mounted on the gimbal takes image of the direct under of the UAV's position. This image is processed in real time on the onboard GPU to generate the pixel coordinates of the helipad in the image. The generated pixel coordinates are transferred to the onboard FCC to generate an outer loop controller command, which is input to a commercial controller that performs attitude control. As a result, we confirmed that UAV can be successfully landed on the helipad, which is the desired landing position, by using the system constructed through the flight experiment