Towards Seismic Profile Similarity Based On Multi-feature Fusion

With the development of the oil industry and computer technology, it is a research trend that the image processing algorithm is used to analyze the seismic profile and it has become a research hotspot that the seismic profile similarity is used to classify geological body.The seismic profile is a typical natural texture and texture feature extraction is the basis and key process of similarity analysis. Therefore, texture feature extraction algorithm has become the focus of the dissertation. Firstly, towards the seismic profile texture feature and the effectiveness of the texture feature extraction method, the texture feature algorithm is studied based on the HLBP extraction method. Secondly, since HLBP extraction method has certain problems as it has threshold with strong subjectivity and poor locality and also the extracted texture details and edges are unclear, a multiple feature fusion based extraction algorithm is proposed. The Gaussian weighted matrix is introduced when the HLBP feature is extracting. Then the adaptive threshold and fusion feature can be extracted. Finally, the validity, reasonableness and description ability of proposed algorithm are verified by compared with the LBP and HLBP. By using histogram intersection and chi-square statistic to measure the histograms of fusion features separately, the inner similarity of seismic profiles and the outer similarity of natural texture images are analyzed.The experimental results show that the fusion algorithm can effectively avoid the influence of the manually configured threshold. Furthermore, the fusion feature can accurately describe the texture feature of image. Thus, a data foundation is laid for the similarity analysis of seismic profiles and natural texture images.