Study On Noise Removal For Airborne Electromagnetic Data Based On Principal Component Analysis And Minimum Noise Fraction

Airborne electromagnetic exploration technology is a geophysical survey method based on electromagnetic induction theory,which is widely used in geological survey,mineral exploration,water resources exploration and natural environment monitoring.The noise of airborne electromagnetic exploration is characterized by multiple sources and complexity,which seriously affects the fine exploration ability of the homemade airborne electromagnetic exploration system for deep underground anomalous bodies.Removing the complex noise of airborne electromagnetic data is very important for the breakthrough of airborne electromagnetic system in China.By analyzing the time domain and frequency domain characteristics,the time-frequency characteristics and the statistical characteristics of the time domain airborne electromagnetic data and its noise,we search the denoising method suitable for airborne electromagnetic profile data.Airborne electromagnetic profile data and noise are partially overlapped in the frequency domain,filtering methods in time domain or frequency domain can not denoise effectively.The correlation between signals in airborne electromagnetic profile is high,and signals are not related to noise.The noise is Gauss,and the statistical characteristic analysis method can be used to denoise.In order to solve the above problems,the principal component analysis(PCA)denoising method is applied to denoise the airborne electromagnetic profile data,the airborne electromagnetic profile data is linearly transformed into a set of principal components arranged according to the variance,and the low order principal components representing the signal are reconstructed to the electromagnetic data to remove nosie.In order to solve the problem that the low order principal component contains a lot of noise which can not be removed when the noise variance of the profile data is greater than the signal,the minimum noise fraction(MNF)is used to remove the noise of the profile data.The data is linearly transformed into MNF components arranged according to SNR,and the noise is removed by reconstructing low order MNF components with high SNR.But when the noise variance is smaller than the signal variance,the minimum noise fraction denoising result will remain partial small noise.Combined with the advantages of principal component analysis and minimum noise fraction,a MNF filtered principal components reconstruction denoising method is proposed.The principal component analysis is used to remove the small uncorrelated noise in the airborne electromagnetic profile data,and then the minimum noise fraction is used to filter the large variance noise in the low order principal components,and the low order principal components after the filtering is reconstructed to denoise.The simulation and field data de-noising results show that this method can not only effectively remove the noise of the profile data,but also enhance the resolution of the airborne electromagnetic exploration to the abnormity of the underground.