GigaPan Image-Based 3D Reconstruction for Engineering Geological Investigation

Image based 3-dimensional (3D) reconstruction is a powerful technique for quantifying the shape of complex objects, and interactive gigapixel panoramic photographs provide multi-scale imagery that facilitate unprecedented long-range qualitative observations. This research examines approaches for combining these technologies with the goal of generating 3D reconstructions from multiple ultra-long range gigapixel panoramas. Because restricted access to the geological object is often an issue for a close-range observation. The overall workflow involves: (1) acquiring multiple overlapping gigapixel panoramas from various viewpoints providing a triangulation angle of approximately 5°; (2) camera calibration; and (3) creating a 3D model using specifically adapted Structure-from-Motion and subsequent Multi-View-Stereo techniques. Initial studies performed at a dolomite quarry located near Graz, Austria, show that a dense and accurate 3D point cloud can be obtained at a range of 630 m. Rock discontinuity orientations measured directly from the model are in excellent agreement with traditional compass measurements, model dimensions (lengths and slope angles) agree with actual physical quantities, and geologic features can be mapped and projected into the 3D model in very high detail. These initial results indicate the developed methodology can greatly assist engineering geologists making quantitative measurements and analyses of outcrops and exposures that are inaccessible or at such great distances so as to preclude traditional photogrammetric techniques. To better define the limits of application, ongoing research is focused on evaluating range effects, in particular the maximum range (and associated parameters) to which this methodology can be reasonably applied in the context of engineering geological investigations.