LIRA - Landmine Intelligent Rescue Autonomous System

Abstract

The continued global threat of unexploded landmines and other unexploded ordnance dictates the necessity for the design of safe, efficient, and cost-effective detection technologies. This paper presents a new autonomous robotic solution to overcome critical limitations of prior systems, primarily their high false-positive rates and high operational costs. This new model couples a low-cost metal detector using a microcontroller with a state-of-the-art computer vision system powered by the YOLOv11n deep learning algorithm running on a Raspberry Pi 4, combining them into a hybrid approach allowing for a unique pause and analyze workflow. The rover autonomously stops when detecting a metallic signature and then uses visual analysis to validate the presence of a target object be it mine, person, or animal. The system performance is demonstrated through a comprehensive evaluation on a held out dataset, reaching 90% precision and 92.3% recall, strongly reducing spurious halts compared to sensor only methods. This work showcases an easy to build project utilizing widely available hardware with a robust rocker bogie mobility platform for a practical and scalable solution for humanitarian demining and other hazardous reconnaissance applications.