With the development of complex velocity structure inversion and improvements in high-performance computational power, it has become possible to calculate accurate 3D strain Green tensors (SGT) by means of high-resolution 3D earth models. Due to more uniform distribution and better point source effects, small-to-moderate earthquakes are widely used in local-scale waveform-based tomography. As it is known, waveform-based tomography methods like wave-equation tomography and full waveform inversion (FWI) require focal mechanism solutions during the whole inversion procedure. However, traditional focal mechanism inversion methods are not very suitable to small-to-moderate earthquakes. In this study, we first used the finite differential method to construct a strain Green tensor (SGT) database and filtered the synthetic and actual waveforms in different frequency band ranges. We then relocated the earthquakes by minimizing the cross-correlation travel time differences, and, next, inverted the focal mechanism by minimizing the waveform residuals. Finally, we used this SGT database method to inverse the focal mechanisms of a series of M_W3.4-5.7 earthquakes in the Longmenshan fault zone, between the eastern margin of the Tibetan plateau and the western part of Sichuan basin. The results indicated that the proposed method can be applied in (near) real-time inversions of focal mechanism solutions since the database can be pre-constructed.