Autonomous quadcopter landing system using monocular camera and fractal marker

Multirotor landing systems relying on GPS/INS navigation alone have insufficient accuracy due to the change of geometry of the transmitting satellite or object obstruction to the GPS receiver. In this study, Fractal Marker was added as a visual target to improve the localization accuracy of the multirotor during the landing. The marker can be detected by the onboard camera, which allows for a six-degree-of-freedom (DOF) pose estimation by utilizing the onboard computer. This computer also acts as a high-level processor to control the quadcopter during the landing by using a simple velocity-limited proportional control to set its velocity on horizontal axes and a fixed setpoint velocity for the vertical axis. The test results show that the proposed method reduces the landing positional error by 88.3% compared to a non-camera landing while showing a high level of reliability. Furthermore, the landing trajectory and landing time are analyzed to understand the tradeoffs between the two tested methods