First arrival pickup method based on strong noise and weak signal detection

Abstract

The invention discloses a first-arrival pickup method based on strong noise and weak signal detection, and provides a first-arrival pickup process for mountainous area data, aiming at the mountainous area data with complex surface conditions in China, the signal-to-noise ratio is poor, the inter-channel time difference changes drastically, and a good pickup result cannot be obtained by a traditional pickup method. The method comprises the following steps: firstly, carrying out a preprocessing process, including inter-channel time difference jump suppression based on a smooth floating reference surface, amplitude compensation, first arrival wave filtering by an anisotropic Gaussian filter and the like, then carrying out high-dimensional and multi-attribute extraction, then determining an approximate range of first arrival pickup through a clustering algorithm, and finally carrying out multi-attribute Markov decision first arrival pickup, and a final pickup result is obtained. Compared with a traditional method, the method realizes automatic and intelligent pickup, and aims at seismic data with low signal-to-noise ratio and violent inter-channel time difference change in mountainous areas in China.

Description

technical field [0001] The invention belongs to the technical field of earth exploration, and in particular relates to a first-arrival picking method based on strong noise and weak signal detection. Background technique [0002] The early first arrival wave picking was manually picked by experts based on experience. The process was very time-consuming and relied too much on experience. Later, interactive and partially automated picking methods appeared, which also required a lot of time for manual quality control. With the application of high-density seismic acquisition technology, the amount of seismic data has become very large, resulting in a sharp increase in labor costs during the first arrival picking process. , The intelligent first-arrival picking method has become the focus of attention. At present, the algorithms used for first arrival picking mainly include: energy ratio method (Coppens, 1985; etc.), AIC (Maeda, 1985; etc.), higher order statistics (HOS) (Saragio...

AI Technical Summary

Problems solved by technology

The method based on the instantaneous characteristics of the earthquake is sensitive to noise, and it is difficult to accurately determine the first arrival position under the condition of low signal-to-noise ratio

The method based on the overall characteristics of seismic records has a certain inhibitory effect on noise, but it is not suitable for complex mountainous data.

The neural network method requires a large number of samples to improve the training accuracy, and the implementation process is complicated and lacks generalization ability

Based on reinforcement learning theory, Ma et al. (2019) regarded travel time picking as a Markov decision process, and automated global optimization on the energy ratio spectrum to achieve first arrival travel time picking, but this method lacks detailed information on reward function and initial state selection. description, difficult to adapt to complex waveforms

[0003] With the application of high-density seismic acquisition technology, the amount of seismic data is increasing, and the automatic and intelligent first-arrival picking method has become the focus of attention. For seismic records with simple surface conditions, the general first-arrival picking algorithm can obtain better results. However, for seismic data with low signal-to-noise ratio and severe inter-trace time difference caused by complex near-surface conditions, especially some piedmont data in my country with developed noise, weak energy, and serious absorption and attenuation, the conventional picking method will It is difficult to meet the demand. The existing methods may focus on a single attribute. Here we combine multiple attribute extractions and integrate multiple attributes into the Markov statistical decision framework for picking

Images