Near-surface geophysical exploration makes extensive use of remote sensing and aeromagnetic exploration techniques due to their noninvasive benefits. However, in geophysical research, the issue of nonuniqueness in a single data source makes it challenging to adequately portray the geological structure. Although multisource detection data fusion can successfully make up for this shortcoming, data fusion is hampered by issues, such as accuracy bias, altitude variations, datum drift, and scale mismatch, because of variations in observation techniques, measurement precision, and sensor coverage. This research suggests a multisource magnetic data fusion method (XGBOOST-HASM) that combines high-accuracy surface modeling (HASM) with extreme gradient boosting (XGBOOST) to solve the aforementioned issues. First, satellite magnetic data were preprocessed using continuation computation and interpolation techniques in order to standardize the height difference and spatial resolution of the observation platform of multisource magnetic data. The XGBOOST model is then trained on the overlapping portions of multisource data with the aim to forecast the satellite magnetic data in the nonoverlapping parts. Finally, input the UAV magnetic sample data and the projected data into the HASM to achieve the best surface fitting. Results from experiments conducted in China's Xinjiang region demonstrate that XGBOOST-HASM creates high-precision magnetic field data with a single datum and the same dimension by leveraging the complementary nature of multisource data. It also shows notable improvements in data accuracy and consistency. Compared with three advanced multisource magnetic data fusion methods, XGBOOST-HASM improves the peak signal-to-noise ratio by 24.3327 dB and reduces the root-mean-square error by 88.53%.
Xu et al. (Thu,) studied this question.