Optimization: Drone-Operated Metal Detection Based on Machine Learning and PID Controller

Abstract

This paper proposes a methodology to detect metals using a drone equipped with a metal detector and programmed by machine learning (ML) models. Our proposed research process could be considered a safe and efficient unmanned mine detection technology for the eventual removal of landmines. Users of this methodology can remotely control the drones without entering the minefield to detect the metal buried and to distinguish whether the metal is mine or not. To realize this idea, we have first stabilized and improved the attitude control of a drone with an attached metal detector by using the micro genetic algorithm-based optimization of proportional-integral-differential control gains. Next, for metal detection, ML models such as a support vector machine and a back-propagation neural network were trained using the annotated dataset. Finally, we have built a controlled drone equipped with a metal detector and trained ML models and experimentally validated our methodology. According to the experimental results, the present study secured the flight stability of the unmanned metal detection drones and the high detection success rate.