Studv On Noise Removal Of Time-doinain Airborne Eleclromamietic Data

Time domain airborne electromagnetic detection is a kind of geophysicalprospecting method, which is based on the electromagnetic induction law and avariety of vehicles as the carrier. It is high efficiency and can explore greatexploration area, etc. Time domain airborne electromagnetic detection, because of itsflight measurement way, introduces a lot of noise, including coil motor noise, thehumanities, instrument internal noise, geological noise, and system noise which iscaused by the environment, temperature change and so on. So data quality and resultsof data processing and image interpretation are seriously affected. As a result, the datanoise removal and repression is the important technical foundation of time domainairborne electromagnetic detection.This article rely on sub-topic of the national863major projects,“study on thetheory and design of system of the time-domain helicopter airborne electromagneticexploration research and practice”, and sub-topic of the special research on the majornational scientific research equipment,“airborne transient electromagnetic systemdata processing and inversion imaging”. This paper study the noise removing methodaccording to characters of atmospheric noise in the raw data, noise of survey pointbefore channeling and profiles noise separately. The mainly content of the thesis andthe research results are as follows:Sferic noise in raw data is short duration, large amplitude and strong random intime domain. At deferent time, different locations, different measurement component,the effects of sferic noise also has bigger difference. So first the sferic noise isidentified and marketed. If the content of noise is less, a pruning method is proposed,otherwise Alpha-trim filtering method. When pruning method is used, data thatcontains sferic noise and the other half of the cycle corresponding to the data areabandoned in stacking. Alpha-trim method is used along the survey line.Based on the EMD algorithm, this paper studies the noise removal of surveypoint data. When EMD is used, the survey point data is decomposed into intrinsicmode functions which are from high to low frequency and a remainder that isrepresentative of the trend. Then the remainder is as the filtered signal. According tothe EMD filter characteristics, high frequency noise included in the intrinsic modefunctions is removed.In view of the residual noise in profiles data, this paper study profiles filteringand irrelevant noise removal respectively based on adaptive window width filteringalgorithm and principal component analysis algorithm. Adaptive window width filteralgorithm adjust the filter window width according to the rate of change of localgradient. In abnormal place, a smaller window is used and otherwise bigger. So the abnormal value is kept and noise is removed. Principal component analysis algorithmconvert the profiles data into principal components. Useful information is contained ina minority of low-order principal component and most advanced principalcomponents contain irrelative noise. Profiles data is reconstructed by a few low orderprincipal components. Irrelative noise is removed.Through comparison of noising removing effect and signal-to-noise ratio, theeffectiveness of the algorithm is proved.