Noise reduction of aerial gamma-ray surveys

Abstract

Two methods of reducing noise in aerial survey data have been evaluated as to their performance on a set of spectra measured with a laboratory spectrometer and a section of an aerial survey. The two methods were Noise Adjusted Singular Value Decomposition (NASVD) and Maximum Noise Fraction (MNF). These are applied to the spectral data before subsequent processing to determine potassium, uranium and thorium concentrations. The MNF method was found most effective at removing noise from U and Th signals, reducing noise in U by a factor of 3.5 and Th by 1.5. No technique investigated gave an effective noise reduction for K. The MNF technique requires that a measure of the noise be obtained and this is best done using the differences of adjoining spectra along flight lines. Parameters such as the number of spectral channels to use in the analysis and the number of MNF components to retain in reconstructing the smoothed spectra were investigated. We recommend keeping around 40 MNF components to ensure minor spatially related signal is retained and using spectral data in the range 200 to 3000 keV.