Research On 3D Seismic Full Horizon Automatic Tracking Method Based On Multi-information Guidance

Seismic horizon tracking is a technique for extracting profile events in post-stack seismic data images,which is an important technique for post-stack seismic data analysis.Based on the generated seismic horizon map,it is possible to predict the reservoir of underground resources and provide a reference for resource exploitation,thereby saving manpower and material resources,and has high application value.The horizon tracking method of 3D seismic data are studied in the thesis,meanwhile,3D horizon is the specific spatial distribution of the 3D seismic data waveform.At present,the mainstream horizon tracking technique is interpreting manually to realize the event axis tracking of two-dimensional seismic data.However two-dimensional horizon cannot reflect the distribution of seismic data in three-dimensional space and the manual tracking method is inefficient and costly which obviously is not a better choice.The purpose of this thesis is to explore the intelligent 3D horizon tracking method,and the expected goal is to improve the quality and efficiency of horizon tracking effect.The main work of this thesis includes the following two aspects.(1)The application of clustering methods in the field of horizon tracking are studied in the thesis,combined with the density clustering method in machine learning,the extreme value distribution and attribute characteristics of 3D seismic data,a 3D full horizon density based on multi-attribute guidance is designed.Clustering tracking method,this method turns the horizon tracking process into the extreme value clustering process of 3D seismic data.On the basis of the clustering method,combining the physical characteristics of actual seismic data,a brand-new 3D horizon tracking scheme is given.In this thesis,a horizon tracking method based on DBSCAN is designed based on the characteristics of the horizontal continuous distribution of extreme points in space.Considering the rich information contained in the actual seismic data,the seismic data is decomposed by the structural feature vector,and the dip angle and fault attribute volume of the seismic data are obtained,which are combined with the three-dimensional horizon tracking algorithm to obtain the multi-attribute guided DBSCAN full horizon.Tracking method,this method mainly compares the continuity of horizons,whether there is a skipped horizon,and the degree of horizon integrity.The 3D horizon tracking method is fast and accurate,the 3D horizon is real and reliable,and it is suitable for extreme value event clustering.(2)In order to realize the full horizon tracking of seismic data sampling points,the algorithm does not depend on the spatial distribution of extreme points,and designs a 3D full horizon ripple diffusion tracking method guided by waveform similarity.DTW is a time dynamic warping algorithm commonly used in speech time series matching.This thesis improves on the basis of DTW,making it an algorithm suitable for calculating the connection relationship between seismic traces.After using DTW to obtain the optimal connection path of the seed points of the seismic trace,select the candidate mapping points within the specific range of the mapping points,and use the diffusion window to calculate the global optimal connection scheme of the seed points,so as to realize the diffusion of horizon labels in the three-dimensional space.That is,corrugated diffusion.A supplementary diffusion algorithm is designed for the case of missing horizons.The algorithm has made some progress in the problem of full horizon tracking of 3D data,and achieved a good horizon tracking effect on real seismic data with relatively fuzzy events,which proves the practicability of the algorithm,and the extracted horizons are complete and continuous.