Images of the velocity cross--section at 10 different depths in Yunnan Province have been suc-cessfully reconstructed, using the principle and method of seismic tomography (ST). These velocitycoss-sections are consistent with the results of six velocity profiles of deep seismic sounding. The reconstruction method of ST used in this paper has two advantages: (i) The velocity modeli determined by spe cifying node--velocities of hexahedron and linear interpolation through knownbasic function. A velocity field with lateral continuous velocity value and vertical discontinuitiesis established. This model is obviously better than the block model of constant velocity. The verticaldiscontinuities allow the model to simulate velocity discontinuities in the crust and upper mantle. (ii)The velocity and hypocenter parameters are separated by introducing an orthogonal projection oper-ator. Sequential orthogonal triangularization and modified singular decomposition are adopted.These measures reduce the RAM requirement Images of the velocity cross-section at 10 different depths in Yunnan Province have been suc-cessfully reconstructed, using the principle and method of seismic tomography (ST). These velocitycoss-sections are consistent with the results of six velocity profiles of deep seismic sounding. The reconstruction method of ST used in this paper has two advantages: (i) The velocity modeli determined by spe cifying node - velocities of hexahedron and linear interpolation known known. A velocity field with lateral continuous velocity value and vertical discontinuitiesis established . This model is obviously better than the block model of constant velocity. The verticaldiscontinuities allow the model to simulate velocity discontinuities in the crust and upper mantle. (Ii) The velocity and hypocenter parameters are separated by introducing an orthogonal projection oper-ator. Sequential orthogonal triangularization and modified singular decomposition are adopted. These measures reduce t he RAM requirement