Stady On Airborne Transient Electromagnetic Method Of Denoising Based On Kernel Minimum Noise Fraction

Airborne transient electromagnetic method(ATEM),as an airborne geophysical survey method based on aircraft,is widely used in mineral exploration,oil and gas evaluation,Marine monitoring and other fields due to its high detection efficiency,low cost and wide exploration area.Due to the detection method of aviation transient electromagnetic system,electromagnetic data is susceptible to various kinds of noises in the actual acquisition process.If the noise in the data is not removed,the accuracy of data processing and inversion will be seriously affected.Therefore,it is of great significance to find effective de-noising methods to remove noise from data.In this paper,the method of airborne transient electromagnetic data de-noising is studied in combination with the sub-topic "Research on aviation ATEM and ground data processing methods and software development" of the national key research and development project "research on integrated processing and interpretation methods of airborne geophysical data and software development".An airborne transient electromagnetic de-noising method based on kernel minimum noise fraction is proposed.By deducing the calculation process of KMNF,the mechanism of de-noising is understood.Aiming at the problem of noise estimation in KMNF decomposition process,this paper adopts adaptive variable window width filtering algorithm to estimate and calculate noise,and verifies its feasibility by model calculation.The kernel function selected in this paper is gaussian radial basis kernel function.Some gaussian noise is added to the forward data,and the linear minimum noise fraction and kernel minimum noise fraction are respectively used for de-noising.Experimental results show that both linear minimum noise fraction and kernel minimum noise fraction can separate gaussian noise and achieve the purpose of de-noising,and kernel minimum noise fraction has better de-noising effect.Power frequency interference has a great impact on data.Through the comparative analysis of adding harmonic interference and adding gaussian noise and harmonic interference de-noising at the same time,the linear minimum noise fraction can remove most of the noise,but at the same time,it loses the useful information of late trace and cannot guarantee the amplitude of abnormal signal.Kernel minimum noise fraction uses kernel transform to decompose data into more components,which can effectively separate useful signals from noise and achieve the purpose of de-noising.Finally,by taking an airborne transient electromagnetic profile data as an example,the kernel minimum noise separation is used to de-noising the data,and good results are obtained,which fully demonstrates the practicability of this method.