lp norm inverse spectral decomposition and its multi-sparsity fusion interpretation.

Spectral decomposition has been widely used in the detection and identification of underground anomalous features (such as faults, river channels, and karst caves). However, the conventional spectral decomposition method is restrained by the window function, and hence, it mostly has low time-frequency focusing and resolution, thereby hampering the fine interpretation of seismic targets. To solve this problem, we investigated the sparse inverse spectral decomposition constrained by the lp norm (0 < p ≤ 1). Using a numerical model, we demonstrated the higher time-frequency resolution of this method and its capability for improving the seismic interpretation for thin layers. Moreover, given the actual underground geology that can be often complex, we further propose a p-norm constrained inverse spectral attribute interpretation method based on multiresolution time-frequency feature fusion. By comprehensively analyzing the time-frequency spectrum results constrained by the different p-norms, we can obtain more refined interpretation results than those obtained by the traditional strategy, which incorporates a single norm constraint. Finally, the proposed strategy was applied to the processing and interpretation of actual three-dimensional seismic data for a study area covering about 230 km2 in western China. The results reveal that the surface water system in this area is characterized by stepwise convergence from a higher position in the north (a buried hill) toward the south and by the development of faults. We thus demonstrated that the proposed method has huge application potential in seismic interpretation. [ABSTRACT FROM AUTHOR]