The time-frequency electromagnetic prospecting system has both the characteristics of frequency sounding and transient electromagnetic sounding system, can obtain apparent resistivity both in frequency domain and time domain, therefore, thus, more accurate and credible geoelectric information can be obtained. After analyzing, there are mainly three typical types of noise in time-frequency electromagnetic observed data, they are 50Hz power frequency and its harmonics interference noise, high-frequency random impulse noise and low frequency interference noise. In view of characteristics of these three types of noise and the time-frequency electromagnetic prospecting signal, we adopted frequency domain band stop filtering method remove the 50 Hz power frequency and its harmonics interference noise, proposed a segmentation and extension median filtering method and a fitting fixed extreme EMD method to remove the high frequency random impulse noise and the low frequency interference noise separately, and proposed a median filtering window size test selection method based on variance and skewness coefficient of sample signal. Moreover, we designed a denoising processing fl ow as handle 50Hz power frequency and its harmonics interference noise at fist step, handle high frequency random impulse noise at second step, handle low frequency interference noise at last step, it can give full advantages of the denoising handling methods. By theoretical analysis and verification by practical observation data experiments, the methods what we adopted and proposed and the processing flow we designed can effectively remove the three types of interference noises effectively and avoid method error, can preserve phase and amplitude information of effective signal accurately, finally, high quality and satisfactory time domain signal waveform can obtained as a result, and lays a good foundation for extracting accurate transient electromagnetic attenuation curve in time domain for the Time-Frequency electromagnetic prospecting data.
Read full abstract