Research On Noise Suppression For Airborne Electromagnetic Data Based On Statistical Characteristics Analysis

Helicopter airborne electromagnetic detection is a kind of airborne electromagnetic detection method based on the plane transport, airborne electromagnetic detection based on Faraday’s law of electromagnetic induction. It has advantages such as high efficiency, wide detection area and the low cost, which is widely used in mineral resources exploration, oil and gas exploration and other fields, especially for the bad natural environment places such as swamps, forest and desert where the ground investigation person cannot work. Because the transmitter coil and receive coil of helicopter time domain airborne electromagnetic system adopt the central loop method, hanging at the bottom of the aircraft. Due to flight conditions, flight speed, temperature, and coil swing and so on which seriously affect the quality of the observation data and identification and interpretation of geophysical information, especially for the deep target. The noise suppression of data processing started relatively late, and have a certain gap compared with foreign countries.With the increasing demand for the mine resources rapidly in recent years, airborne electromagnetic detection technology has been widely concerned. In terms of noise suppression of airborne electromagnetic data, suppression method research for classified noise are primary. The main processing methods are time- frequency filtering, wavelet signal processing technology and etc.. Since the frequency spectrum of noise is overlapped with electromagnetic signal, airborne electromagnetic profile data still exist residual noise after conventional noise suppress ion processing.This article relies on National major research equipment projects sub-topics “study of the ATEM system data processing and explanation software technology”. To solve the above problem, this paper put forward noise suppression technology for airborne electromagnetic data based on statistical characteristics analysis. By studying the principal component analysis, principal component filtering reconstruction and minimum noise fraction technology, the residual noise of airborne electromagnetic data are effectively suppressed.The major findings and content are as follows:(1) Due to the geological information of the earth is slowly changing continuously, there is a certain correlation between channel and channel of airborne electromagnetic data. Based on the theory of multivariate statistical analysis, there is a certain amount of redundant information in the measured observations. For residual noise on airborne electromagnetic data profiles, the noise suppression method for airborne electromagnetic data is proposed based on principal component analysis. Through research for the internal relationship between the covariance matrix and correlation coefficient matrix of observation data, original data are composed into principal components ordered according to the variance on the principal component domain. Eigenvectors of covariance matrix are the directions of the converted principal components. In this way, the transformed principal component not only retain the original information of airborne electromagnetic data, but also are unrelated to each other. The low-order principal component mainly contain airborne electromagnetic data of large variance and higher-order principal component represent noise of smaller variance. The low-order principal component are adopted to reconstruct airborne electromagnetic data to suppress the part of residual unrelated noise. The denoising results of simulation examples and field example verify the validity of the principal component analysis in suppressing residual noise.(2) The principal component analysis method converts airborne electromagnetic data into the principal component to reconstruct, but low-order principal component participating in the reconstruction still contain the high frequency spacial noise which affects reconstructed electromagnetic data. So on the basis of principal component analysis, this paper designed adaptive filtering width low-pass filter group to filter out high frequency spacial noise of each principal component data profile. According to second order difference characteristics of each principal component profile, filtering width of each corresponding point is calculated. The adaptive filtering width low-pass filter group are found to filter out principal component profile. The denoising result of simulation examples and field example verify the validity of the principal component analysis in suppressing residual noise.(3) Due to principal component are ordered according to the variance when the noise variance is larger, the principal component analysis cannot effectively remove the noise. Therefore, the minimum noise fraction is introduced into the airborne electromagnetic data denoising. Minimum noise fraction transformed data into minimum noise fraction component according to signal to noise ratio. It makes the components containing unit variance of noise and uncorrelated each other, so the process of decomposition is not affected by noise. This chapter studied the characteristics of the minimum noise fraction component, characteristic of noise fraction and the relationship between noise fraction parameter and signal-to- noise ratio. Through the analysis of two groups of simulation examples and the field data and the noise suppression results, it verifies the effectiveness of the minimum noise fraction algorithm.(4) This paper adapted empirical mode function method, residual noise method based the theory of multiple linear regression, the autocorrelation fraction method to estimate the noise covariance matrix and studied the influence of the noise covariance matrix estimation for the minimum noise fraction denoising results. Through designing simulation model with deep anomaly, the denoising result of the latest channel and analysis of capacity on reflecting the deep underground abnormal, it verified the three noise covariance estimation methods can effectly estimate the noise covariance matrix.The main innovation lies as follow:(1) Based on principal component analysis, this paper constructed noise suppression method for the airborne electromagnetic data based on the principal component filtering method. Based on the adaptive window wide filter algorithm, according to second-order characteristics of each survey point in principal component profile, it can change the band width of the filter adaptively, filtering high frequency spacial noise on lower-order principal component participating in reconstruction and acquiring more accurate interpretation of later channels.(2) Noise suppression algorithm was proposed based on the minimum noise fraction methods for airborne electromagnetic data. This paper studied noise estimation algorithm and analyzed minimum noise fraction and the minimum noise fraction composition profiles. Through reconstruction of low-order minimum noise fraction components, it solved the problem that principal component and principal component filter ing cannot effectively remove the residual noise when noise variance is larger. Minimum noise fraction method can effectively suppress the noise and obtain the useful signal.