3D reconstruction and characterization of soil desiccation cracks using smartphone photogrammetry method

Abstract

Quantitative measurement of geometric parameters of desiccation cracks is essential for evaluating their development status and potential impacts on soil properties. However, current mainstream digital image analysis methods mainly focus on providing 2D morphological parameters of cracks. To address this issue and provide a useful tool for 3D characterization of desiccation cracks, this study explores the potential of using smartphone photogrammetry. Three naturally developed cracks and five artificially excavated cracks of varying sizes were selected for image acquisition and 3D reconstruction using the photogrammetry method. The depth of reconstructed natural crack models were compared with manually measured depth values to assess the accuracy of this method. Furthermore, an evaluation was conducted by comparing reconstructed artificial crack models through photogrammetry with laser-scanned models. The results revealed that when there is good visibility inside the cracks, the photogrammetry method can effectively characterize desiccation cracks. However, the presence of occlusions or relatively small viewing angle may result in incomplete reconstruction and potential underestimation of crack depth. The root mean square error (RMSE) between the photogrammetry models and the laser-scanned models of the five artificial cracks were less than 0.18 mm. The swelling and disintegration behaviors of the crack networks were effectively captured through smartphone photogrammetry during the field application test. The RMSE for the crack depth measurements extracted from the photogrammetry model was determined to be 0.86 cm. This proposed method has advantages such as low cost and high efficiency, making it a promising practical approach for characterizing desiccation cracks at varying scales.