基于密度模型稀疏表征的重力反演方法

Gravity inversion is an essential tool for imaging subsurface density distribution, and the selection of a proper density model constraining method is key for improving gravity inversion resolution and reliability. Conventional constraining methods match inversion target via adjusting smoothing or sparse constrain weight starting from density model in the mesh space. However, it is difficult to guarantee the reasonability and validity of model constrains when there exists a set of multi-style geologic bodies, an inaccurate anomaly separation or an unreasonable mesh-generation solution. To this end, a novel gravity inversion method has been proposed which is based on sparse representation of density model. Firstly, the representation of the density model to be recovered is assumed to be the linear combination of model feature matrix and sparse decomposition coefficient, then objective function of gravity inversion is re-derived and sparse solving process of decomposition coefficient also is provided. Comparing with traditional gravity inversion methods, the feature model for constructing model feature matrix contains prior geometrical information of multi-style geologic bodies, and the sparsity of decomposition coefficient ensures that the inversion target comes from combination of the most typical geological modes. Finally, the test of synthetic models and the application of field data demonstrate the validity of our proposed gravity inversion method which is based on sparse representation of density model.