top of page

The advantages of Magnetic Anomaly Detection (MAD) and Artificial Intelligence (AI) for marine asset

An article by PhD (c) Leda Tzannetou, AI/ML Engineer at SOTIRIA Technology

The detection of a security risk in an offshore marine environment is a complex problem due to 1) the harsh environmental conditions, 2) the possible line-of-sight absence and 3) the increasing sea noise levels which set limitations to the range and accuracy of existing acoustic techniques.

The magnetic anomaly detection technique has the advantage that water is practically transparent to the static magnetic field, which is not affected by bad weather conditions such as fog, rain, snow and ice. In addition, magnetic sensing is a passive technique, which means that such a detection system can remain undetected, unlike non-passive detection systems such as radars.

MAD algorithms work with data obtained from measurements of earth’s magnetic field using arrays of high sensitivity magnetometers, such as total field optically pumped magnetometers or 2D fluxgate magnetometers. They can detect and track a potential marine security risk even when the magnetic signal is very weak or buried in noise of comparable power. The method detects anomalies in the geomagnetic field, due to the presence and movement of magnetic objects. The algorithms assume that the magnetic fields due to stationary or moving objects follow the Biot-Savart law and process the magnetometer data using orthogonal base functions, signal decomposition techniques and detectors to locate and track the object. The MAD algorithms do not rely on Artificial Intelligence (AI) techniques but can be complemented by machine learning algorithms as data accumulates, such as Density-based spatial clustering of applications with noise (DBSCAN) or other clustering algorithms, to further enhance their accuracy.

AI/ML powered MAD algorithms are being developed by SOTIRIA Technology and fused with acoustic detection models for tailored marine assets and terrain security applications.

bottom of page