Semi-Supervised Learning Based De-Raining Method for UAV

In recent years, unmanned aerial vehicles (UAVs) have made remarkable progress and are highly expected to play an active role in a variety of fields. However, there are still many concerns regarding the autonomous flight and safety of UAVs. Among them, video image degradation due to rainy weather is a significant problem, regardless the UAV is flying autonomously or remotely controlled. In this work, we propose an efficient learning-based de-raining method using video images of UAVs in rainy conditions. The proposed method creates a de-raining model appropriate for the situation by adding UAV rain images to Syn2Real training data. As a result, the proposed method performs better than the existing de-raining method in PSNR and SSIM.