A newly developed imaging system is presented, which measures three-dimensional (3D) deformations of a soil surface in geotechnical experiments involving physical modelling. The method adopts the computer vision technique ‘structure from motion and multi-view stereo’ delivered by an open-source software MicMac. Three, 2 megapixel industrial cameras were synchronised and used to capture images of a deforming soil surface. The images were used to reconstruct the observed scene to a high-density, accurate 3D point cloud. A new method has been developed to process the obtained 3D point clouds and images to determine the 3D displacement vectors. The procedure is highly automatic which allows large data sets to be processed with minimal manual intervention. Two series of quantification experiments were carried out to assess the performance of the system which has shown the overall accuracy to be within 0·05 mm over a field of view of 500 × 250 mm. An example application is presented to demonstrate the capabilities of the 3D imaging system.