Landmine Classification and Prediction Using Machine Learning Techniques

This study aims to classify and detect five common types of landmines using machine learning models based on sensor data and soil characteristics. Landmines pose a severe threat to human safety, particularly in war zones and post-conflict areas. Traditional demining methods are often costly and dangerous. Therefore, data-driven approaches leveraging machine learning can significantly improve the automation and accuracy of landmine detection. In this research, voltage, height, and soil type collected from sensors are used as input features, while the landmine category serves as the target variable. Several supervised learning models, including Support Vector Machine (SVM), Decision Tree, and K-Nearest Neighbors (KNN), are trained to classify the landmine types. Among them, the SVM model achieves the best performance, with an optimized accuracy of 82% and an AUC score of 0.93, significantly outperforming other models. Future work may focus on multimodal sensor fusion, lightweight deployment, and improving generalization in complex terrains to enhance model applicability and reliability.