In this project, my work is to develop a vision system that detects and tracks a water target, on which the deck pattern of a navy warship is placed. The platform we used is a 200 lb Yamaha RMAX, which is equipped with a laser altimeter, a sonar, an IMU (gyroscope and accelerometer), and a camera. My system takes all these sensor outputs and estimates the position and the velocity of the target expressed in world coordinate. A small AprilTag is installed on an expected landing location so that the vehicle can see the target until the last minute of landing, and also the AprilTag can provide the attitude of the target, which is critical to landing on a water target. The vision system runs in real time and provides continuous states of the target to the navigation system. The vision system uses an extended Kalman filter to estimate the velocity via the observed positions and to fuse the outputs from multiple vision techniques (which are AprilTag and deck pattern detection). In the simulation, vision-based closed-loop landing (which means the navigation is completely dependent on vision) is successful, and also a flight test for that scenario is ready. In the flight test movie above, we could not achieve landing due to a malfunction of the boat, but we were able to confirm that the vision system was working with real on-board images. In the next flight test, a vision-based closed-loop landing should be achieved.