Enhancing Anti-Interference of Magnetic Tracking: A MagRobustNet-Based Framework With Self-Supervised Anomaly Detection and Measurements Recovery

Magnetic tracking technology shows great promise for applications in medicine and industry. However, it often suffers from diverse and unpredictable interferences in practical applications, such as hard-/soft-iron interferences and sensor saturation, leading to reduced localization accuracy or even tracking failure. Thus, we propose a two-step framework to mitigate the impact of interferences based on MagRobustNet, a UNet-like autoencoder network. In the first step, disjoint mask sets are used in conjunction with MagRobustNet to detect anomalous measurements subject to disturbances. In the second step, the interfered regions are masked, and MagRobustNet is applied again to recover their expected measurements from neighboring normal data. Experimental results from testing in four interference scenarios showed that the proposed method improved the average position accuracy by 76.2%, enhancing the tracking system's anti-interference capability. In addition, the proposed method can indicate the interfered regions, thereby prompting the adjustment of the magnetometer array to an interference-free location and offering a new potential diagnostic method for localizing ingested foreign bodies in clinical practice.