This paper introduces an ensemble learning-based approach aimed at enhancing the accuracy of landmine detection while significantly reducing false alarms. The proposed methodology integrates three ensemble learning techniques, namely Bagging, Boosting, and Stacking. Each of these techniques leverages three machine learning models: a Convolutional Neural Network (CNN), a Support Vector Machine (SVM), and a Random Forest (RF). The performance of the proposed models was assessed using key evaluation metrics, including accuracy, recall, confusion matrix, F1-score, and ROC curve. These evaluation tools facilitate a comprehensive analysis of the detection system’s robustness and optimize the mitigation of classification anomalies. The primary objective of this approach is to enhance detection capabilities while minimizing errors, thereby strengthening safety measures in potentially hazardous areas.