The Identification Of Quasi Similar Noise By The Radial Trace Transform And Its Attenuation In Seismic Exploration

Seismic exploration is an important exploration method in the coal、oil and gasexploration. It is difficult to avoid the various noise sources in the seismic explorationprocess. And these various noise sources caused the diversity of nature of noise in theseismic records. These noises have been problems in recognizing and tracking thereflection events to get the subsurface information. These random noises often lowerthe signal-to-noise ratio (SNR) and make the data processing difficult. A main task ofseismic exploration is to remove random noises in seismic data so as to improve itsSNR. In Seismic data processing, we choose different filtering methods for thedifferent kinds of noise attenuation. The quasi similar noise proposed in this paper isone of the distortional random noises. This noise is quasi similar to the short reflectionevent because it shares similar spatial characteristics with the short reflection event.For example, its spatial coherences are quasi similar to the short reflection event froma certain angle.This similar nature leads to many difficulties in separating these twokinds of information. Aiming at this problem, we introduce the idea of removing thequasi similar noise after its identification. In this paper, we adopt the idea of radialtrace transform (RTT) to differentiate the quasi similar noise and use thetime-frequency peaking filter (TFPF) as a tool to remove the noise.In this paper, we improve the radial model based on the radial trace transformtheory, using the local fan of common-origin radial traces for radial trace transform.Through the RTT, the original signals would be rearranged from the coordinates x tto the new coordinates v t. The spatial relationship of the original signals would bechanged in the new coordinates.The advantages of the radial trace transform inrearranging the spatial characteristics would break the the similar nature of the twokinds of information. Therefore, the quasi similar noise can be recognized from thedifferent performances before and after the RTT. We can separate the two kinds ofinformation by the TFPF with different window length.We have done lots of experiments according to the algorithm proposed in thispaper. Both experiments on the synthetic model and the field data prove the goodperformance of this method in the two kinds of information identification. Throughthe radial trace transform, the short reflection event has reduced dominant frequency,while the quasi similar noise has no obvious changes in the frequency domain.Therefore, the quasi similar noise would be recognized according to the differentperformances before and after the RTT. The results are basically consistent with thethe theoretical results and illustrate that the method works well on seismic records. After the identification of the quasi similar noise, we use the TFPF with long windowfor the noise and short window for the short reflect event. Both experiments on thesynthetic model and the field data prove the good performance of this method inseparation of the two kinds of information. In a word, the method proposed in thisletter that removing the quasi similar noise after its identification can separate the twokinds of information.