DEEP LEARNING BASED SEMI-SUPERVISED FRAMEWORK FOR SAFE LANDING OF UNMANNED AERIAL VEHICLES

Abstract

Nowadays, various remote sensing systems have adopted the use of Unmanned Aircraft Systems (UAVs) due to the increasing popularity of drones. Hence, there is a need to develop such methods for the safe landing of aerial drones for suitable landing site identification. This research focuses on developing a Machine Learning (ML) based framework for identifying suitable places to ensure the safe landing of drones. An Autonomous Safe Landing of Aerial Drone Vehicles (ASLAD) framework, a novel and innovative approach, was proposed in this regard. This framework was based on the Self-training Semi-supervised Learning Model (SSLM). For experiments, high-resolution aerial images were taken. All the images were resized using image resizing techniques. Image splitting technique was used to split the dataset into labeled and unlabeled classes. Experiments showed that the proposed model produced promising results with an accuracy of 96.75% and an F1-score of 94.45%. The results showed that the proposed framework can identify suitable places for the safe landing of aerial drones.