Vision-Based Autonomous Landing Solution of UAVs in Dimming Light Environment

The performance of vision-based autonomous landing schemes of unmanned aerial vehicles (UAVs) highly depends on the surrounding light condition. Low light environment usually results in poor landing accuracy of UAVs. In order to enable precise landing under low light conditions, in this paper, we propose a systematic solution that combines target detection, state estimation, and landing control. For target detection, a deep learning approach is employed to compensate for dim-light image feedback. In addition, a particular landing sign is customized to aid flight state estimation in visual adverse conditions. The estimated state is forwarded to the flight controller for precise landing control. In order to evaluate the performance of the proposed method, case studies of both static and moving landing signs were simulated in Gazebo. As a result, the performance of the proposed method is not only at per with the common solutions such as ArUco Library under normal light conditions, but it can also work properly under low light conditions that other methods cannot.