ABSTRACT
A study by Le et al. (2017) reported the application of computer vision techniques structure from motion (SfM) and multi-view stereo (MVS) to measure three-dimensional soil displacements at the surface of physical models. However, little information exists on the significance of the camera resolution and the number of images to the measurement performance. This study assesses the measurement performance of the SfM-MVS, provided by an open source software Micmac, with input images taken by two different types of camera including DSLR (18Mega-pixel) and mobile phone cameras (12Mega-pixel). Rigorous quantifications were carried out to examine the precision of the image analysis, in measuring vertical and horizontal displacements, over a region of interest of 420×200 mm. The measurement precision, achieved by different numbers of images, ranged from 0.06 mm to 0.03 mm. The results from this paper can be useful for researchers to select appropriate camera that satisfies their measurement requirements.
