The Research Of Spatiotemporal2-D Time-Frequency Peak Filtering Based On Matching Pursuit

Seismic exploration is one of the most effective methods in oil and gas resourceexploration. But due to the varied terrain, complex subsurface structure and the increase ofdetecting depth, resulting in the signal-to-noise ratio of the seismic data is very low. Thenoise will seriously influence the accuracy of seismic exploration. In order to increase thesignal-to-noise ratio of seismic data under the strong background noise has extremelyimportant meaning, and also it is very hot in the research on seismic data processing.Time-Frequency Peak Filtering can effectively suppress the random noise in lowsignal-to-noise ratio seismic data. The condition of Time-Frequency Peak Filtering tosuppress the random noise is the signal must be linear. The real seismic signal is a nonlinearsignal, so in order to increase the linearity of the seismic signal for the Time-FrequencyPeak Filtering, we adopt the seismic signal with a time window to make the signal locallinearization, but for some high frequency seismic signal, even if the window length whichis very short also can’t achieve linear conditions, so to improve the signal’s linearity for theTime-Frequency Peak Filtering is the focus of this paper.In this paper, combined with the seismic reflection events is bend, we put forward amethod that using spatiotemporal2-D filtering trace re-sampling the seismic signal toimprove the linearity of the seismic signal for the Time-Frequency Peak Filtering signal.First we select the spatiotemporal2-D filtering trace which the bending degree is close tothe bending degree of the reflection events to resample the original seismic signal. Becauseeach channel of the seismic signal has a very strong coherence, the seismic signal’samplitude after resampling is similar, the linear degree of seismic signal is high, so theparallel Spatiotemporal2-D Trace Time-Frequency Peak Filtering can satisfy the linearitycondition. But the real seismic records contain many reflection events which have differentbending degree, so the main problem is to get the spatiotemporal2-D filtering trace.In order to match the spatiotemporal2-D trace of the reflection events, we use theMatching Pursuit method to preliminary identify the seismic reflection event trace, and thento fit filtering trace. Firstly, decompose the seismic signal into the signal area and noise area,through the matching process we can get the time parameters of each wavelet, and then use the wavelet time parameters and the seismic channel number to get the spatiotemporal2-Dtrace for Time-Frequency Peak Filtering; then use the spatiotemporal2-D trace forspatiotemporal2-D Time-Frequency Peak Filtering to resample the seismic signal, so theseismic signal for the Time-Frequency Peak Filtering is linearity, so it can recover theeffective signal better.In order to verify the applicability of this method, first we apply it to the syntheticseismic data and the real data. The experimental results show that this method can match thereflection event very well and increase the seismic signal linearity. So this method canrecover effective signal from the low signal-to-noise ratio seismic data, suppress the randomnoise, improve the continuity of the reflection events to make the low signal-to-noise ratioseismic data more coherent, more clearly.