HawkEye Drone UAV for advanced landmine detection and autonomous control

Abstract

Landmines continue to pose a severe threat to civilian populations in post-conflict regions, where conventional detection methods are often slow, risky, and ineffective against modern non-metallic mines. This paper presents an autonomous UAV-based landmine detection system that relies solely on thermal imaging and machine learning techniques. A lightweight MobileNet-V3 convolutional neural network (CNN) is employed to analyze thermal signatures captured by an onboard infrared camera and classify potential landmine locations in real time. A mixed dataset, composed of publicly available thermal landmine images and custom-collected field data, was used for training and validation. Experimental results demonstrate reliable detection of buried non-metallic objects under varying environmental and soil conditions. The system supports autonomous navigation, GPS-based mapping, and Geofencerestricted flight using ArduPilot, enabling safe aerial surveying of hazardous regions. The proposed approach highlights the effectiveness of deep learning-based thermal analysis for contactless, low-risk humanitarian demining operations.