Abstract: Aiming at the low resolution problem of smooth inversion in gravity exploration, the authors proposed a local smooth constrained three-dimensional inversion algorithm based on certain prior information such as the depth of geological bodies and stratum inclination, and a storage method of roughness matrix in smooth inversion. By this method the computer's calculation memory can be reduced through storing the M×N dimension roughness matrix as M×2 dimension, and when the roughness matrix is multiplied by other matrices in this storage mode, the position information stored in the roughness matrix can be read,and the calculating results can be obtained. The theoretical model and measured data were used to perform inversion trial calculations using the algorithm proposed in this study. It is shown that the result of the local smooth inversion algorithm is more accurate than the result of the global smooth inversion, and the algorithm is still stable,effective,and feasible under a given noise level in production.

Key words: gravity gradient, roughness matrix storage, partial smooth constraint, inversion