An improved principal component analysis approach to noise removal for airborne electromagnetic data

Abstract

Summary There is residual noise in airborne electromagnetic data even after pre-processing by routine time and frequency domain de-noising techniques. It affects the quality of later channels seriously. The principal component analysis (PCA) approach could suppress the residual noise by low-order principal components reconstruction. It removes noise in principal component (PC) domain other than time and frequency domain and is effective especially to those having the overlap spectrum with the airborne electromagnetic data. However, the low-order PCs still contain residual noise which behaves as spatial high frequency noise along the PC profiles. Therefore the adaptive-width smoothing filter is adopted to remove the spatial noise along the low-order PC profiles. The filtered low-order PCs are then used to rebuild the airborne electromagnetic data. The improved PCA approach not only deletes the uncorrelated noise containing in the high-order PCs but also rejects the spacial noise in the low-order PCs. The CDI results for synthetic data have proven that the improved PCA get better conductivity interpretation for the airborne electromagnetic survey.