The absolute liquid surface inspection uses a liquid metal plane as the basis for calibrating the flatness of optical plane crystal. The high reflectivity of liquid metal leads to an inability to match light intensity, making it challenging to obtain high-quality interference images. To address this issue, a method of attenuating light intensity by using a metal mesh has been adopted. This method reduces the light intensity by exploiting the diffraction effect of the metal mesh and achieve intensity matching. Further analysis on the selection of the mesh number of the metal mesh and how to suppress the introduced error has also been conducted. Derive the relationship between zero-order diffraction efficiency and the number of sieve mesh, and verify it experimentally in a Zygo GPI 4-inch interferometer. The results show that the selection range for the mesh count of the screen is between 120 and 200. In the experiment, a 150-mesh sieve was selected, yielding high-quality interference patterns. An uneven distribution of the metal screen mesh can introduce wavefront errors. In the simulation process, a random wire error in a fixed direction was given, the resulting wavefront errors are distributed in a grid shape. These are mid-to-high frequency errors that can be suppressed using low-pass filtering. The liquid metal calibration experiment was carried out with a 300mm aperture Fizeau interferometer. The face shape is consistent with the static liquid face shape distribution, the PV value deviation is 0.007λ, and the RMS value deviation is 0.0036λ. It is compared with the measurement results of water and 0.65Cst silicon oil as the liquid surface, and the three face shape distributions are basically the same. The experiment proves that the use of a metal screen can effectively solve the problem of high reflectivity of liquid metal, and the introduced error can be suppressed by using a low-pass filter.