Research On 3D Horizon Automatic Tracking Method Based On Robust Estimation Of Slope Attribute

Horizons can be regarded as geological time synchronous surfaces with specific ages,and horizon tracking is to mark the geological synchronous surfaces in seismic data using color coding and other methods.Based on the extracted horizon calibration map,it is possible to reveal ancient sedimentary environments,structural geological models,and can also be used for stratigraphic interpretation and exploration of oil and gas resources,which has high research and application value.The original 3D seismic data is typically generated by stimulating artificial sources to produce seismic waves,which encounter geological synchronous surfaces and result in reflections and refractions,and are received by detectors arranged on the ground.These data usually contain complex noise,which increases the difficulty of horizon tracking work.Most existing horizon tracking technologies rely on manual methods on 2D seismic data profiles.However,2D horizons cannot accurately represent the actual underground geological situation,lack information on the distribution of geological horizons,and manual tracking is costly and subjective.Therefore,this thesis aims to study a high-noise robust 3D seismic horizon automatic tracking method.The goal is to obtain complete and accurate horizon tracking results from low signal-to-noise ratio seismic images.The main innovations of this thesis are as follows:(1)The slope attribute of seismic data can be used to guide horizon tracking work.In response to the problems of low signal-to-noise ratio of original seismic data,low resolution of traditional slope estimation methods,and significant impact of noise,this thesis explores the application of multiscale methods in the field of seismic slope estimation in image processing.By using multiscale methods to deal with low signal-tonoise ratio data,a compromise between resolution and accuracy can be achieved.Based on the gradient structure tensor algorithm,a 3D seismic image slope estimation method based on multiscale gradient structure tensor is proposed.This thesis extends the Gaussian image pyramid to 3D space,and calculates the slope estimation and corresponding quality measurement of different scale layer images based on the gradient structure tensor algorithm.Based on quality measurement,a fusion mechanism for 3D image pyramids is proposed to achieve multiscale slope estimation fusion and ultimately obtain the target slope.This method is applied to both artificial and actual seismic data in work areas.Compared with traditional algorithms,the slope estimation results have no significant decrease in resolution,high accuracy,and fewer outliers,which can effectively guide subsequent horizon tracking work.(2)Traditional horizon tracking methods mostly rely on manual operations on 2D cross sections,lacking global consideration.Some 3D methods also fail to produce satisfactory results when dealing with cross-fault tracking.This thesis proposes a 3D multi-horizon tracking method based on slope attribute guidance.The algorithm fits the target horizon’s slope with the locally estimated local slope from the multi-scale gradient structure tensor algorithm,and uses quality metrics as weighting factors to balance the unstable values in the fault-discontinuous region.The model is then extended to multiple horizons,and vertical constraints are introduced to regulate the relationship between the horizons.The final horizon positions are obtained through multiple iterations.By applying this algorithm to data from multiple real work areas,compared with the comparative algorithm,the tracked horizons are complete and smooth,follow the background trend,accurately cross discontinuous areas such as faults,and significantly reduce the phenomenon of cascading horizons.The algorithm results are outstanding.