Study Of Horizon Tracking Method In Three-Dimensinal Seismic Images Based On Multilevel Frame

With the development of artificial intelligence technology, feature classification technology has become an important form of data analysis. It has been widely used in data mining, artificial intelligence, pattern recognition and other fields. Feature classification technology identify and analyze the relationship and categories of data to produce a model, and then use this model to analyze the new data features to determine the category of new data. This articles combine feature classification technology with feature of horizon in the three-dimensional seismic data. And design three-dimensional horizon tracking methods based on this theory.The three-dimensional seismic images are distribution of seismic waveform in three-dimensional space. Horizon tracking is to get the horizon information in the three-dimensional seismic images. Currently, horizon information in the three-dimensional seismic images is obtained by artificial method. The geologists analyze every two-dimensional profile and get the horizon information in the two-dimensional profile. Because two-dimensional profile can’t contain all the information of the three-dimensional seismic images and the geological structure exist in the way of three-dimensional, so the artificial method has serious deficiencies in accuracy and efficiency. The three-dimensional horizon tracking method analyses the three-dimensional seismic images directly and in a global view. So it has great superiority in accuracy and efficiency. This article is response to these problems, and the main work consists of two aspects:Firstly, horizon characteristic and classification technology have been studied. We combine waveform characteristics in three-dimensional seismic image, spatial distribution characteristic of extreme points and classification technology, and design the horizon tracking method based on multilevel frame. This method analyzes horizon tracking question from extreme point level, layer level, and expert-level. In extreme point level, horizon extreme points connection algorithm has been designed based on study of lateral continuity of horizon extreme points in three-dimensional space. And this method connects the extreme points into horizon fragment. In layer level, Horizon fragment merge algorithm has been designed based on study of waveform similarity and GMM clustering algorithm. And this method merges horizon fragments into big horizon. In expert-level, in order to make geological interpretation can adjustment the result according to the actual demand, horizon correction algorithm has been designed. And this method makes the horizon tracking results more realistic geological structure. The horizon tracking method based on multilevel frame analysis horizon tracking question from extreme point level and achieve automatic tracking horizon information in three-dimensional seismic images.Secondly, this article extracts horizon features method extracts features in the three-dimensional seismic images from a new point of view. And design the horizon tracking method based on match search by studying vertical distribution characteristic of horizon in the three-dimensional seismic images. Horizons are tracked individually in traditional tracking methods. So relationship between different horizons has been ignored. Horizons present a layered distribution in the three-dimensional seismic images. There is a gap between adjacent layers and the gaps between different layers have some differences. These are the vertical distribution characteristics of horizon in the three-dimensional seismic images. The horizon tracking method based on match search is based on these characteristics, and contains vertical distribution characteristics extraction algorithm, block generating algorithm based on match search and block connection algorithm based on amplitude guiding. Vertical distribution characteristics extraction algorithm and block generating algorithm based on match search connect the extreme points into horizon fragment, and block connection algorithm based on amplitude guiding merges horizon fragments into big horizon. Scarcity in the horizon handled lastly. Because this method takes advantage of relationship between the layers and can track horizon in the three-dimensional seismic images concurrently. So the efficiency and accuracy have been improved.