Advanced Machine Learning Detection using UAV Motion Tracking and Control

Drones, or unmanned aerial vehicles (UAVs), have become more commonplace thanks to the booming commercial UAV sector. The potential for these devices to do significant harm, whether on purpose or by accident, has prompted urgent security concerns. In recent years, several solutions have been presented by academics and businesses to the problem of protecting vital infrastructure. Because to its superior resilience, computer vision is often employed as an autonomous drone detection solution over other offered alternatives such as RADAR, acoustics, and RF signal analysis. As deep learning algorithms are among the most efficient of these computer vision-based methods, they are increasingly being used in practise. In this work, we offer a self-sufficient drone tracking and identification system that makes use of a stationary beam profile and a lesser camera on a movable turret. In order to make the most optimal use of both resources, we suggest a combined multi-frame machine learning detection approach, where the frame from the zoomed turret camera is superimposed on the frame from the wide-angle static camera. With this method, we are capable of building a cost-effective pipeline that allows for the simultaneous identification of tiny sized aerial invaders on both the main picture plane and the magnified image plane. In addition, we provide the whole system, which includes tracking techniques, deep learning classification frameworks, and protocols.