To solve the problems of traditional snow depth measurement methods, such as large investment, time consuming, and low efficiency, a contactless method named Photogrammetry was proposed based on SfM (structure from motion). First, a series of experiments were conducted using the wind-snow combined experiment facility, and snow distribution photos were obtained by camera. Then, the three-dimensional coordinates of the snow surface were restored, and the spatial distribution data of the snow surface was extracted by MATLAB programming, which was compared with the snow depths that were measured by steel ruler to verify the accuracy of the method. The results showed that the accuracy was kept at ±2 mm and the relative error was within 1% when using photogrammetry for simple roofs. While when using photogrammetry for complex roofs, the accuracy was ±4 mm and the relative error was within 1%. It was proved that the accuracy of the photogrammetry method satisfied the measurement requirements of the tested model and the efficiency could be greatly improved when using photogrammetry for measuring snow depth distribution on complex test model roofs.