| Fault and anticline etc. are significant features in geological structure, which function crucially in the transfer and accumulation of the oil-gas, and lead to complex features in seismic section. As the deepening in the exploration extent, the macro-structured and easy-to-find oil and gas reservoirs have continuously been developed. Those clearly-formed and large-scaled structures become less and less. How to find the replacement area of reserves and how to develop the remaining oil-gas have already been the major problems in the development of exploration. Therefore, the focus has been shifted to the hard-to-find subtle traps, especially the exploration of the low-amplitude oil and gas reservoir.At present, there is not yet an integrated and rigorous definition concerning the low-amplitude structure in the theories and practices of the oil-gas exploration both domestically and abroad. It is a relative concept, which indicates the geological body of smooth structure and low amplitude about 10m to 50m. Although it is of low amplitude, it is easy to gather the oil-gas, and form the"tiny and rich"oil and gas reservoir. Due to its low amplitude, straightness in the reflection event and small varying amplitude are reflected in the seismic materials, with the structure hard to identify.The edge of the image is one of the most fundamental and most significant features, which is the aggregation of the spatial-mutated pixels generated by the gray scale. The purpose of the identification and extraction of the edge is to highlight the edge information of the image and heighten the contour feature of the image.Through the detection of the images by use of every detection operator and the comparison of the results, it is discovered that due to the lack of smoothing processing, without strong ability to resist the noise, Roberts operator's test result is relatively rough and loses a number of initial image information, though it can exactly locate the edge; compared with the Sobel operator, Prewitt operator includes weight in the average process of filtering, but it can not detect the image completely under the interference of noise; Log detection operator aims to find the precipitous zero cross point through the calculation of second order differential. The detection result of Log operator is much more superior to the previous two methods. Canny detection operator utilizes first order differential to unveil the edge point of the image. Comparing with Roberts, Sobel and Prewitt, Canny method owns high effects in de-noising and detection, and can better detect the complete edge of the target image.When the amplitude values of every sampling point are transformed into various levels of grey scale, the picture of seismic section becomes the picture of seismic grey scale. Thus the image edge detection can be introduced into the treatment of seismic section. Then the reflection event turns into the edge of two-dimensional numerical image. Through identifying the events in the practical seismic section by use of every operator, extracting the envelope medial axis from the identification results, it is discovered that the identification results of the events based on the Roberts operator, Prewitt operator and Sobel operator are relatively poor, and lead to a great loss in the original information and an obvious decrease in event number, and even ruptures. While the identification results of events of Log operator and Canny operator are comparatively better, which almost reserve all the horizon information of the events in the original seismic section. However, there are still a few tiny spotted noise in the detection result section, which due to the appearance of the false boundary point caused by the discontinuous grey scale. Compare the results after filtering the noise, Canny operator possesses the best detection effects, more complete horizon information and continuous event.Coherence analysis is a technology based on the computation of the similarities of the data among seismic traces, in order to get the horizon information and stratum continuity, which functions greatly in the identification of underground small structures, faults and cracks. The mathematical theoretical foundation of C1 coherence algorithm is normalized cross correlation. Through the treatment of waveform similarities among seismic traces, C1 algorithm is easy to realize and without much computational complexity. It shows high coherent value in continuous and transversely uniform stratum, and shows low coherent value in case of faults and cracks.The realization method of C1 coherence algorithm is as follows: 1.Open a fixed time window with appropriate size in the data of the reference trace and the surrounding seismic traces which need to be computed, according to the time shift characteristic of coherence, fix the reference trace, take the interval of the sample points as the unit, and move the adjacent traces from the top down. With every movement, calculate the coherent value of the waveform in the time window once. 2. Figure out the maximum value C1(Max) , that is, the value of position when the two traces are mostly correlated, in order to demonstrate the two traces own the maximum similarity in this position. 3. Take down the time-shift value in accordance with the maximum coherent value, so as to reflect the continuous transverse change of the wave form in horizon, i.e., the rise and fall situation of the structures.First build a horizontal reflection interface and a tilted reflection interface. Through the detection of C1 coherence algorithm, record the searching range, the value ofτcorresponding with the C1 (Max), in the purpose of certifying the horizon trend of the events. And take the time depth of C1 (Max) as the ordinate, and the seismic trace as the abscissa, and get the tracing graph of reflection horizon. After the evaluation of the inaccuracy and comparison of the results, it is discovered that the result acquired is in complete uniformity with the theory, which verifies the practicality of C1 coherence method. Build respectively three kinds of synthetic seismogram of low-amplitude structured model. They are H=50m(λ/2),L=100m(λ),H/L=1/2;H=10m(λ/10),L=50m(λ/2),H/L=1/5;and H=5m(λ/10),L=50m(λ/2),H/L=1/10. The features presented in the seismic section are concealed structure, straight-distributed event, and with difficulty in ensuring the existence of the horizon and the size of the structure. After C1 coherence identification, from the horizon tracing graph the horizontal position of the low-amplitude structure can be obviously seen. Combining with the horizon slope rate graph, it can be ascertained exactly the time depth of the structure's end point and its high point, and its corresponding seismic trace. Therefore, complete spatial display information of the low-amplitude structure is acquired.The major factors influencing the C1 algorithm's identification of low-amplitude structure are searching range, noise and the size of time window. Build a model of low-amplitude structure of two layers and one interface. The three low-amplitude structures H=50m(λ/2),L=150m(3λ/2),H/L=1/3,H=20m(λ/5),L=60m(3λ/5),H/L=1/3,H=30m(3λ/10),L=60m(3λ/5),H/L=1/2 are distributed on this reflection interface continuously. The searching range begins to increase fromτ=1dt, and different identification results are gained whenτ=4dt andτ=15dt. Comparing with the case whenτ=1dt, it is discovered that S3 needs larger searching range than S1 and S2. From this, it can be seen that the parameter influencing the searching range is the height-width ratio instead of the size of the structure's amplitude. The fix of searching rangeτis decided by the height-width ratio of the real horizon of the structure, which is the maximum inclining angle of the target horizon in the work area. Add Gaussian independent noise of 10dB and 0dB respectively to the model of low-amplitude structure, and then use C1 coherence algorithm to identify, it is found that the identification result of 10dB is without any mistake, and the identification result of 0dB appears to be no more than four thousandth relative error, which would not affect the fix of the structural horizon. Therefore, under the environment of Gaussian independent noise, the identification result of C1 coherence algorithm towards low-amplitude structure is still reliable. Take 5T*/2～T*/4 as the time window to identify the low-amplitude structure under the environment of 0dB. When the time window is 3T*/2～T*/2, a complete wave crest or wave trough could be seen, which is comparatively ideal and can reflect the relatively correct spatial display of the low-amplitude structure. When the time window is less than T*/2, the view of the time window is comparatively narrow, and a complete wave crest or wave trough can not be seen. The result at this moment does not demonstrate the similarities of waves among seismic traces, but rather the noise situation of this small area. When the time window is more than 3T*/2, on the one hand, the amount of data included by the time window is enlarged, the time of calculation would be prolonged; on the other hand, the oversize time window would not be so sensitive towards the identification of the low-amplitude structure. In addition, the oversize time window would encompass other horizon information if other horizons exist in the stratum. Therefore, the value of the time window should be 3T*/2～T*/2Build a model of low-amplitude structure of underlying interference with three layers and two interfaces. Take the layer thickness D=λ,λ/2,λ/4,λ/5 to form the seismic records containing 10dB's noise. After the C1 coherence identification, it is discovered that when the layer thickness is reduced, within the limit of appropriate time window, the proper reduction in the width of the window can heighten the identification effects and reduce the inaccuracy of identification. When the interference layer is contained in the noise, C1 coherence algorithm still can acquire the horizon spatial display of low-amplitude structure correctly. Apply C1 coherence algorithm to the actual identification of seismic section, and the identification results reflect the event trend of reflection wave of the target horizon T1, T2 and T3 comparatively exactly. The effect of application is relatively ideal.Complex trace analysis is three-instantaneous analysis, which aims to get the physical parameters reflecting the characteristics of seismic wave, such as the instantaneous amplitude, the instantaneous phase and the instantaneous frequency from the signal of the seismic trace. As the increase in exploration extent, the complex seismic trace can not meet the need of the exploration of the thin layers. Therefore, the complex trace analysis of high resolution is derived after the derivation of the signal, which is weighted through the spectrum of the signal x (t), taking ?ω2as the weights.This paper builds a wedge model, through complex trace analysis of high resolution, which raises the resolution limit of the conventional seism fromλ/4 toλ/5, and functions positively to the amplitude. Do the complex trace analysis of high resolution to the model of low-amplitude structure of underlying interference with three layers and two interfaces and try to get its three-instantaneous section of high resolution. This shows the horizon of the low-amplitude structure, which can not be displayed in the original section, in the seismic section of high resolution, and fixes the existence of the structural horizon in the instantaneous amplitude of high resolution through the amplitude highlight. It also confirms the time depth of the low-amplitude structure in the instantaneous phase section of high resolution, and the end point of the structure in the instantaneous frequency section of high resolution. To identify the seismic section of high resolution by the use of C1 coherence algorithm, compared with the identification result which have not been analyzed by the complex trace analysis of high resolution, it is discovered that the identification result erases the false structure of underlying horizon, and the selected points are almost completely coincide with the true values, with the selected relative error decreasing from 0.8% to 0.2%, which is an conspicuous enhancement in the identification result and as well beneficial to the determination of the spatial display of the low-amplitude structure.
|
PDF Full Text Request