During the fieldwork, we will be implementing some new archaeological techniques to do our research. One of these new techniques is Computer Vision Photogrammetry for underwater archaeological site recording. Computer Vision Photogrammetry allows us to load a series of overlapping pictures of the shipwreck into dedicated software in order to automatically generate an accurate three-dimensional model.
The process is as follows. Divers film the site with a GoPro in a lawnmower pattern, thereby ensuring overlap. The video footage is then loaded into a program that takes still pictures from the footage. Finally the pictures are then loaded into software called Agisoft Photoscan.
Agisoft Photoscan uses a so-called ‘feature detection algorithm’ to automatically identify and match features in overlapping pictures. Based on the detected features and the camera calibration parameters, Photoscan aligns the pictures relevant to one another. The result is a so called sparse point cloud, which is a 3D approximation of the scene in the pictures. The software now knows the original camera positions and the camera calibration. With this information it can calculate where the pictures overlap. The additional feature points that are then created, are added to the existing sparse point cloud. This creates a much more detailed dense point cloud.
Using the dense point cloud Photoscan creates a surface mesh from the dense point cloud.
Finally Photoscan ‘imposes’ parts of the original pictures onto their corresponding points in the surface mesh.
Computer Vision Photogrammetry reduces underwater recording time and produces an accurate, detailed and objective three-dimensional result. The created model does not only look good, but can also be used for measurements and further research. The pictures in this blog are off course just a small part of the ship. We are looking forward to the final results and will keep you updated! Check our blog tomorrow for more information on the diving techniques used in this campaign.
Robert de Hoop (intern Maritime Program RCE)
With special thanks to Thomas van Damme