Application Of DBSCAN Algorithm In Seismic Facies Classification

Along with complex degree of oil-gas exploratory development increasing and seismic interpretation technology being more and more mature day by day, seismic oil-gas forecasting technology develops toward fine and the practical direction. It requires people to continuously improve the level of awareness, use scientific methods to understand and master the unknown status of existing oil and gas, and explore more new seismic information to predict oil and gas from existing geophysical, geological, extract information such as reservoir development and to. Currently one of the most effective methods to interpretation of seismic data is seismic facies analysis.With the continuous development of seismic acquisition, seismic profile contains more abundant seismic information, which the traditional "physiognomy Law" has not been detected. It has to rely on seismic data processing and computer technology to extraction and analysis. Currently the common seismic data processing technology is K-means algorithm, Fuzzy Clustering Methods and Self-organizing Feature Map Network. The algorithm has some disadvantages: the need to determine the number of clusters by people; difficult to establish accurate and reasonable membership function; computing time is long when amount of data is large and high dimension, sometimes it almost can not be achieved and so on.In response to these problems, we explore the Density-Based Spatial Clustering of Applications with Noise algorithm apply in the seismic facies classification. DBSCAN algorithm does not need to determine the number of cluster by people, computes faster, and can effectively handle noise points and find the spatial clustering of arbitrary shape. Firstly, this paper generates the test data according to the seismic data, and compares DBSCAN algorithm with K-means algorithm, FCM algorithm, SOM clustering analysis algorithm in five areas: the request to input parameters; finding any shape clustering; the processing capacity in massive data, in high dimensional data or with noise data; Then it is applied in the real seismic amplitude attribute data set. The results show that the DBSCAN algorithm can divide seismic facies well and satisfactorily. However, the computing time of this algorithm is a little long, this paper has taken measures of compensate seismic data to make up it, the computing time decreased significantly while guaranteeing the same division effect.