Aeromagnetic Compensation Algorithm Robust to Outliers of Magnetic Sensor Based on Huber Loss Method

Jian Ge,Han Li,Hongpeng Wang,Haobin Dong,Huan Liu,Wenjie Wang,Zhiwen Yuan,Jun Zhu,Haiyang Zhang
DOI: https://doi.org/10.1109/jsen.2019.2907398
IF: 4.3
2019-01-01
IEEE Sensors Journal
Abstract:The outliers of the magnetic sensor that are inevitable in special aeromagnetic surveys reduce the robustness of ordinary least-squares (OLS), which are widely used for aeromagnetic compensation. To address this problem, we propose an aeromagnetic compensation algorithm based on the Huber loss method that is robust to outliers. In the proposed method, different weights are assigned to the inliers and outliers using an iteratively reweighted least-squares technique. Although the OLS performs similarly to the proposed method when only 1% of the data are outliers, it is theoretically verified that the proposed method can increase the goodness-of-fit to 0.9963, from 0.6618 in the case of OLS, in the presence of 10% outliers. An experimental platform was constructed to record real magnetic data, with special measures taken to ensure the presence of outliers in the collected data. The results of a flight test using this experimental platform demonstrate that the proposed method increases the improvement ratio to 4.14 from 2.46 when using the OLS.
What problem does this paper attempt to address?