Federated/Deep Learning in UAV Networks for Wildfire Surveillance

Abstract

The unmanned aerial vehicle network (UAV- net) has been attracting substantial attention as a solution of wildfire surveillance. Application of federated learning (FL) for the UAV-net can provide an applaudable solution to mitigate wildfires. Each UAV can hover at different locations and obtain images with distinctive features. Therefore, it is regarded as an efficient methodology that each UAV fulfills different levels of deep learning (DL) in a distributed and collaborative fashion, which is a new paradigm raised by FL. This paper examines current state-of-the-art research works on detecting wildfire utilizing DL and UAVs. Further, this paper proposes utilizing FL for the UAV-net to monitor and detect wildfire. The impact of different convolutional neural network (CNN) models and layers with tailored model parameters on the performance of prediction accuracy, is addressed with simulations.