In order to further improve the measurement accuracy of 3D laser scanning technology, this paper studies the optimization method of point cloud grid filtering from the perspective of optimization filtering algorithm and improved algorithm based on Kriging, considering the space-related properties of description objects and aiming at the filtering processing problem of point cloud data. Based on engineering practice, grid processing and analysis of 3D laser point cloud data are carried out through field monitoring and comparison test, and the deformation data obtained from the test is compared with the measured data by traditional methods. The results show that the improved algorithm based on Kriging filtering can identify and extract the tunnel contour cross-section's visual data efficiently and obtain the deformation of the tunnel efficiently and accurately. The experimental data of vault subsidence is close to the traditional measurement data, while the peripheral convergence data has some differences. Therefore, tunnel deformation monitoring under 3D laser scanning can better reflect tunnel deformation's real condition under certain environmental conditions and provide a sufficient safety warning for tunnel construction.