Accurate 3D surface measurement of mountain slopes through a fully automated image-based technique

In this paper an automated procedure for surface reconstruction that can be used for surveying and monitoring rock and ground slopes is presented. This method has been developed for geological and engineering applications, where accuracy and completeness are factors of primary importance for the final 3D model, which must provide a detailed metric survey and not only a visual reconstruction of the scene. The proposed procedure integrates two image matching techniques. The first one is used to automatically extract a set of tie points that are needed for computing the exterior orientation parameters of all images through a photogrammetric bundle adjustment. Such tie oints are also exploited to obtain a preliminary seed model that is then enriched based on Multi-photo Least Squares Matching. During this second stage, the surface model is improved in terms of point density and accuracy. Different strategies were implemented to successfully combine both techniques, along with some new improvements. The presented procedure has been tested in two different applications: geometric modelling of rock cliffs and evaluation of weathering of a ground slope. In both cases the obtained results presented accuracy sufficient for geological investigation. Moreover, outcomes were comparable to the ones from laser scanning surveying and other photogrammetric implementations.