Real-time detection and removal of sun glint regions for reliable ocean monitoring using a drone

In this paper, we present an algorithm to determine the sea-surface specular reflection and restore regions affected by scattered glitter for ocean monitoring tasks using unmanned aerial vehicles (UAVs). Unlike land areas, strong sunlight reflection over the sea surface saturates large parts of the images observed by UAVs, and it can hinder the UAVs or operators from performing vision-based missions. To detect a vast sun glint region, we propose a method predicting the occurrence of vast specular reflection by using both solar angular position and the flight information of UAVs. Additionally, to reduce the effect of irregularly scattered glints, we propose an accurate glittering region extraction and restoration method. We evaluate the performance of the proposed approach using various real ocean sunlight reflection images obtained by a UAV. Using the proposed methods, the occurrence of the specular reflection can be effectively detected, and the irregular reflection areas can be extracted and removed in real-time. © ICROS 2019.