Geo-referenced 3D Video as visualization and measurement tool for Cultural Heritage Lazaros Sechidis, Vassilios Tsioukas, Petros Patias The Aristotle University of Thessaloniki, Department of Cadastre Photogrammetry and Cartography Univ. Box 473, GR-54006, Thessaloniki, Greece sechidis@hypernet.hyper.gr , tsioukas@egnatia.ee.auth.gr , patias@topo.auth.gr Keywords: Stereoscopic vision, Epipolar images, 3D video Abstract Just a few years ago, a video card with stereoscopic display capabilities was assumed as a state of the technology product and only professionals had the ability to have one. Rapid technology progress combined with competition among video cards producers, have eliminated this cost so that a typical PC user can afford it. In the near future these cards are expected to be standard output devices of a PC. So, maybe it is the right time to consider of a product that can use this technology: the 3D video. Additionally, if this video includes proper metadata can be a Geo-referenced 3D video. Architects community can use this added valued product for better presentation of their work. Archaeologists can show their sites and findings in a new visual way. And this product gives the ability to make some basic measurements on it. Typical products of a DPS are digital elevation models, orthophotos and/or video using fly-through techniques. With the help of these products, 3D video can be easily created. Once the rendered 3D model has been created, a virtual pair of cameras can capture images of the draped model from any place above or in front of it. Images taken from two different positions away of the object containing the epipolar geometry and can be considered stereo images. In an active polarization system using 3D glasses, these images can be interlaced producing a 3D scene. Additional 3D scenes can be produced from any other position across a virtual path that is drawn above or in front of the object’s model. Finally all the scenes taken along the path defined by the user can be combined and a 3D video is generated. Additionally, the position and the rotation of each camera, for every frame, can be recorded into video file; every frame then equals to a pair of images with known interior & exterior orientation. 1. The concept of Geo-referenced 3D Video One of the early (and still of the most popular) visualization tools has been the Video. Either in uncompressed (eg. AVI) or compressed forms (eg. MPEG) have been extensively used to animate otherwise static screens, or to allow the user to visualize (although not interactively) different parts (and in different scales/zoom factors) of the interest object. Through the time, the Video output proved to posses other useful characteristics as well : (a) It can be used to record and visualize both real scenes and computer generated one; notably both at the same time if so wished, leading to what is called “Augmented Reality”. (b) It is a readily available output of every CAD, photogrammetric or visualization tool. (c) Widely accessible, freeware video players can play it back. (d) It is, by nature and practice, well-integrated in WWW applications, as well as highly compatible to web browsers (eg. VRML). (e) It is the only way to visualize moving objects. Apart of these obvious advantages, Video also provides huge and up to now only partially explored potentials: Construction of stereoscopic video sequences for one, or video “geo-referencing” for another; and many others too. For instance, generally, video construction within the above mentioned tools (CAD etc) is done through the design of a flight-over or walk-through “path” and choice of the “virtual camera” positions, orientation and lens characteristics. That is both the interior and the exterior camera orientation parameters are known. By just shifting this path appropriately to accommodate parallax another path can be generated and thus another video sequence. Both sequences can now be stereoscopically viewed (since their relative orientation on a frame-by-frame basis is known by construction), creating thus a 3D video! Moreover: a “Geo-referenced” 3D video, where every pixel in every video frame could be assigned “real-world” 3D ground coordinates, by simple ray intersection. At another instance, if video is the input of a real scene to be photogrammetrically reconstructed, then each video frame can be “tagged” afterwards (ie. after the photogrammetric processing) by information concerning their position and orientation. This potential has never been used, although this kind of information was always