Observing soil erosion process at fine spatial and temporal scale is of great significance to the study of soil erosion mechanism. A digital close range photogrammetric observation system based on wireless networking technique was explored and established. The evolution of soil surface topography was dynamically monitored by instantaneous image acquisition at different time intervals during ongoing rainfall. Noises on the images such as raindrops was removed by K-means clustering, digital point clouds were calculated and digital elevation model (DEM) was then generated. The results showed that the measurement precision of the established system could reach a sub-millimeter level, and the minimum measurement error was 0.0062mm. The maximum relative error between the measured value and the actual value was -2.9683%. According to the experimental observations, the average relative error of soil loss was -1.73%, and the accuracy of single observation was up to 99.26%. The established digital photogrammetric observation system could accurately calculate the digital point cloud from the underlying surface with 1min time interval and 2mm spatial resolution. The observation methods explored provided a reliable way to monitor soil erosion processes, especially under rainfall conditions, which was of great importance in understanding soil erosion mechanisms.