UAV On-Board Emergency Safe Landing Spot Detection System Combining Classical and Deep Learning-Based Segmentation Methods

摘要

This article proposes the system designed for automatic detection of emergency landing sites for horizontally landing unmanned aerial vehicles (UAV). The presented solution combines the classic computer vision algorithms and novel segmentation methods based on deep learning techniques using the U-Net inspired network architecture. The presented system uses a single nadir camera mounted on a UAV and the energy-efficient compute module capable of highly-parallelized calculations.