Evaluation of field methods for 3-D mapping and 3-D visualization of complex metamorphic structure using multiview stereo terrain models from ground-based photography

Abstract

Field geology has traditionally relied on two-dimensional, paper-based workflows. Although digital mapping techniques are rapidly replacing paper ones, three-dimensional (3-D) terrain models and 3-D visualizations have the potential to revolutionize field studies, yet to date, few studies have embraced this technology. The development of structure-from-motion (SfM) photogrammetry has allowed routine production of high-resolution terrain models from a series of photographs taken at arbitrary angles using "multiview stereo" (MVS) software. However, few studies have applied the MVS approach outside of specific, highly controlled field environments that are easily accessible. In this study, we examine methods for ad hoc application of ground-based MVS in remote field areas with large-scale (>2 km 2 ) multifaceted topography and complex geology. Specifically, we emphasize methods that could be employed in a typical geologic field study without the use of specialized equipment beyond a camera, and we identify various pitfalls that can be avoided during this type of work. We present several scenarios that illustrate the different ways that MVS can be implemented in the field. These scenarios vary with respect to: (1) the manner in which ground control points (GCPs) are collected and distributed; (2) the baseline-to-distance ratio of the imagery; (3) the number of photographs taken; and (4) the type of camera used. Each scenario yields 3-D terrain models from which plane orientations can be extracted and upon which 3-D linework can be drawn. We caution that if absolute accuracy- the difference between the location of the objects on the model and their true position on a geodetic coordinate system-is critical to a project, then great care must be taken in using MVS models obtained solely from ground-based photographs because several factors can contribute to spatial errors as large as hundreds of meters over scales of a few square kilometers. The two primary factors that contribute to these significant spatial errors in MVS models are (1) the distribution and positional accuracy of GCPs and (2) the baseline-to-distance ratio. Nonetheless, MVS is a tool that can easily be applied to any field study regardless of terrain complexity, scale, or accessibility, and it has the potential to revolutionize field studies, particularly in areas with steep terrain.