Autonomous drones based forest surveillance using Faster R-CNN

Abstract

Unmanned Aerial Vehicles (UAVs), often called drones, are airships that do not involve a human pilot or can carry any passengers. UAVs are an element of the Unmanned Aircraft System (UAS). They are accommodated with a ground station and a system to facilitate communication with the drone. UAVs can be controlled by a human operator using a remote control, as remotely-piloted aircraft (RPA), or with differing degrees of independence, from autopilot assistance to fully independent aircraft that does not require human interference. Lately, UAVs have been made capable of flying beyond visual line of sight (BVLOS), due to which autonomous drones have been used in various areas such as commercial, warfare, aerial photography, agriculture and forestry, and law enforcement. Autonomous drones are being used to perform aerial surveillance of forest areas to detect and locate forest fires. Various deep learning algorithms are implemented along with image processing techniques to create a robust mechanism for detecting fire and smoke from the images or video frames obtained from the UAVs. This paper studies and proposes how image processing models and techniques can be incorporated into Unmanned Aerial Vehicle systems to detect smoke and forest fires.