SONAR Based Under Water Mine Detection Using Machine Learning Algorithms

In modern naval warfare, submarines play a vital role in strategic operations. However, submarines face a significant risk of encountering ocean mines, which are lethal and can cause severe damage. To mitigate this risk, submarines are equipped with Sound Navigation and Ranging systems (SONAR) which uses active/passive SONAR to detect objects in the surrounding water. However, distinguishing between the mines and other underwater objects, such as rocks, is a critical challenge. Machine learning based object detection is proposed in this paper to distinguish between under water rock versus mines. techniques. By leveraging the capabilities of machine learning algorithms, submarines can make informed decisions in real-time, helping to ensure the safety of naval forces and successful mission outcomes. Real time dataset of SONAR recordings of mines and rocks is collected and considered for classification. Six machine learning algorithms are considered in this work. Support Vector Machines (SVM), Decision Tree Classifier, Random Forest Classifier, KNN Classifier, XGboost and Ensemble methods are used to grade whether the detected object is rock or mine. Comparative results of all these algorithms are presented in this paper.