PhotoScan Professional

The SfM model processing is performed using Agisoft PhotoScan Professional [50 ]. Processing of the SfM model consists of Agisoft’s standard workflow: alignment of photos to create a sparse cloud, manual identification of GCPs and CPs in images, optimization of camera alignment, generation of a dense point cloud and generation of a 3D textured mesh. Default PhotoScan values are used for processing with a couple of the key settings listed in Table 2. PhotoScan is selected as the primary modeling program based on performance, ubiquity in the literature [51 (link),52 (link),53 (link)] and processing reports. Eventually, the density of the point clouds, the error in location of the GCPs and CPs, and comparisons of time and photos required demonstrates the merit of multi-scale models and methods. Visual comparisons are also made using Bentley ContextCapture [54 ] because of meshing ability and ease of model navigation [55 (link)].
The processed multi-scale SfM model is comparatively analyzed by priority region for accuracy, resolution, visual clarity, and reconstruction quality. GCP and CP error residuals in the SfM model are analyzed using JMP^® [56 ]. Resolution mapping is performed in CloudCompare [57 ] (which is open source) and visual clarity and reconstruction quality are assessed via Acute 3D Viewer [54 ].

To collect complexity metrics, five randomly generated 5 m transects were overlayed on each 3D reconstruction. Then, for each transect, the topographic profile cross-section was calculated (S2 Fig in S1 File). Using these cross-sections and the orthomosaics of reef structure, the following complexity metrics were obtained: a) rugosity index, b) average rate of change in elevation (verticality), c) hard coral cover, d) available refuges, e) feeding field-of-view and f) grazing surface area. Each of these metrics is obtained from five independent (1-D) transects; in effect we were applying traditional liner 1-D measures of complexity within a 3D digital reconstruction. Each metric is described in detail below, all measurements were taken within Agisoft® PhotoScan Professional.

In November 2012, the entire Lotus site was mapped on transparent plastic film (Fig 3, S1 Fig). The tracks of Layers 1 and 2 were measured and photographed for 2D and 3D analyses. The original tracings on plastic film have been reposited at the Qijiang National Geological Park. Replicas were made of several sets using latex for the initial molds and plaster of Paris and fibreglass replicas. They are housed in the Qijiang National Geological Park Museum, with additional replicas in the University of Colorado collections.
The maximum track length (L), maximum width (W), maximum depth (D), pace length (PL), stride length (SL), pace angulation (PA), rotation (R), trackway width (TW) and the angle between digits II–III, III–IV were measured for the ornithopod trackways.
Photogrammetric images were produced from multiple digital photographs (Canon EOS 5D Mark III) which were converted into scaled, highly accurate 3D textured mesh models using Agisoft Photoscan Professional. The mesh models were then imported into Cloud Compare where the models were rendered with accurately scaled color topographic profiles.