퇴적물 원격분류를 위한 개선된 초동신호결정 알고리즘 개발

Abstract

In seismic exploration, the first arrival detection is the task of determining the first signal arrivals as accurately as possible. The accurate detection of the first arrivals needs to calculate the static corrections in a fundamental stage of seismic data processing. For the seismic data with a low signal-to-noise ratio, the first arrival automatic detection method is an important step but it is not easy. Determining of the first arrivals is similar to finding the bathymetry in the ocean exploration. In the set of seismic traces, the method of the first arrival detection is divided in a way directly or using an automated algorithm. When using the former method, users should read the figure for each trace, which takes first arrival travel time. This method is time-consuming; also, the picking error can be relatively large. Therefore, this method is usually both inefficient and unreliable. The latter method aims to pick the first arrival time automatically. It is intend to reduce the picking errors using an automatic picking algorithm. First arrival detection algorithms can be divided into several types of methods, including the coherence method, the neural network method, and the cross-correlation method. Coherence and neural network methods assume the existence of patterns in the first arrivals. Cross-correlation methods are considered to be most appropriate for the near-surface surveys since their algorithms are based on trace-by-trace evaluation of the first arrival times. For first arrival detection, it has been used with a variety of ways of energy ratio (ER), modified energy ratio (MER), and short-term average/long-term average ratio (STA/LTA) such as a cross-correlation method. The energy ratio method is based on sliding time windows; however, the conventional method based on sliding time windows is not stable. The modified energy ratio method was designed to overcome this disadvantage. The peak of the modified energy ratio is much closed to the time of first arrival on noise-free seismograms. The STA/LTA ratio method is similar to Coppens’ algorithm. The difference is to do the ratio of two averages of energy between a short-term window and a long-term window, which is denoted as the STA/LTA, instead of calculating the ratio of energy of seismogram of the two windows in Coppens’ algorithm. The basic idea of this method detects first arrival time when the STA/LTA ratio exceeds a pre-defined threshold. This method is well suited for the first arrival detection, but not for providing accurate arrival times due to the delay associated with the length of the short-term average window. In this study, it is proposed an improved algorithm that can determine first arrivals even in low SNRs signals by combining ER, MER, STA/LTA, exponentially weighted low-pass filter, and Kalman filter precisely. In addition, it is developed and introduced basic signal analysis and processing application which applied to improved algorithm. It is proposed the depth of submarine cable installation, attenuation of the surface related multiple, combining the two types of seismic data, and preliminary result of seabed characterizing based on similarity index using first arrivals of the developed algorithm and application. Key to the effectiveness of the improved algorithm is the manual selection of the first arrival on the initial trace. Thus, the Kalman filter or the exponentially weighted low-pass filter can be used to predict the first arrival time of the next trace in advance. The ER method showed that the first arrival is determined to be approximately 2.0 to 3.0 ms earlier than actual first arrival time due to the noise of the upper water layer. The MER method yielded better results than the ER method. However, the first arrival is determined to be approximately 0.5 ms earlier than actual first arrival time. The STA/LTA method showed more consistent results than the ER method and the MER method. However, the first arrival is determined to be approximately 1.0 ms earlier than actual first arrival time. The improved first arrival detection algorithm showed sound results in finding the first arrival time correctly.