Gaussian beam tomography uses the Gaussian beam operator to compute the travel-time tomographic kernel function instead of the conventional ray tomographic kernels,which can improve the accuracy and stability of velocity building.However,the absence of prior geological information sometimes leads to the lack of geological understanding of the results.On such a basis,by introducing the preconditioned model regularization operator,a method of Gaussian beam tomography velocity building under the constraint of structure-guided filter is derived,whose key is to construct the structure-guided filter operator.In this paper,structure tensors were calculated by seismic profiles,and geological structure features were introduced by anisotropic diffusion smoothing operators as regularization constraints for inversion.This method is entirely data-driven and achieves the "soft constraint" of tomographic inversion,which not only reduces the issue of multiple solutions of inversion but also improves the inversion resolution,and hence can obtain a reasonable velocity model with geological recognition.The validity and practicability of the Gaussian beam tomography velocity building with structure-guided filtering were verified by its application to synthetics and practical data.