Abstract: Digital Image Correlation （DIC） technology enables the measurement of more comprehensive and accurate deformation data, and has been widely adopted in deformation capture in scaled model tests. To obtain the overall deformation trend of an aeolian sand subgrade, a scaled model tests on aeolian sand was conducted and a speckle fabrication method suitable for scaled models within a model box was proposed. This approach extends the applicability of DIC in constrained observation environments. Compared to traditional data acquisition methods, this method significantly enhances the acquisition efficiency of scaled model tests while ensuring the completeness of experimental data. The results show as follow: First, by utilizing pre-installed measures involving a custom detachable model box and high-transparency glass, a speckle fabrication on the observation surface of the model box was achieved, eliminating interference with initial deformation caused by the installation of the high-transparency glass. Secondly, a speckle quality using the Mean Intensity Gradient （MIG） was evaluated and the speckle fabrication method for materials with good water permeability was optimized, such as aeolian sand. The MIG value increased from 6.55 （common white background with black speckles） to 11.19 （black background with white speckles）, and further to 31.46 （white speckles made with acrylic paint）. This significantly enhances speckle contrast, avoiding issues such as overexposure, distortion, interference, and information loss, thereby successfully applying DIC in deformation measurement of the aeolian sand subgrade model. Subsequently, settlement of the main surface of the subgrade model occurred under the fifth load level （100 kPa）. The maximum settlement reached approximately 0.8 mm by the eleventh load level （220 kPa） and approximately 1.3 mm by the sixteenth load level （320 kPa）, at which point loading was terminated due to cracking on the subgrade surface. Finally, the settlement deformation was concentrated near the loading area, with settlement on the near-slope side being greater than that on the inner side of the subgrade. Under the action of a 16th-level load, the maximum settlement of the loading area exceeded 27.67% of the slope side; exceeded 29.36% of the inner side of the roadbed; and the maximum settlement deformation of the slope side exceeded 20.27% of the inner side of the roadbed. This difference was explained from the microscopic perspective of particles being squeezed out laterally. The method of conducting DIC tests using the improved model box exhibits clear systematic nature and portability, providing a complete technical solution for model tests under similar constraints. The revealed deformation results and mechanism analysis of the aeolian sand subgrade can offer certain references for similar experiments and engineering applications.