Rapid and Automatic Reconstruction of Large-scale Areas

点击翻译按钮获取中文摘要

In recent years there has been an increasing demand for realistic virtual worlds representing large-scale, real-world areas comprising of terrain, buildings, trees, cars, etc. Many successful applications employing such realistic virtual representations have already been reported including advanced disaster management simulations for the training of emergency response personnel, visualizing new structures in-situ prior to construction for urban planning and development, and computer games where the storyline takes place in an actual rather than fictional location. For example Ubisoft’s "Assassin’s Creed" contains numerous cultural heritage sites [albeit some of them not in perfect condition], such as the Hagia Sophia, the Galata tower, the castles in Kyrenia and Limassol, etc. However, despite the large volume of work in the area many challenges still remain. Currently, the creation of realistic large-scale 3D content remains a complex, time-consuming, expensive and labor-intensive task. In fact, the creation of models is still widely viewed as a specialized art, requiring personnel with extensive training and experience to produce useful models. Fundamental research is essential in order to bridge the gap between the current state-of-the-art and the ultimate goal of rapid and automatic creation of large-scale areas. This proposal addresses the current technological difficulties of rapid and automatic reconstruction of large scale areas and seeks solutions for the development of accurate, robust and scalable methods and systems for processing the big data captured by active and passive sensors in order to produce a realistic virtual representation. More specifically the research program proposes further study and development of novel and robust algorithms for accurately detecting and extracting: (a) structural information from data captured from passive and active remote sensors i.e. aerial/satellite images and LiDAR, and reconstructing the geometry of the terrain, buildings, cars and tree models representing the acquired area, (b) appearance information from imagery captured from ground, oblique-aerial and satellite sensors, and fusing this information into realistic composite texture atlases of the 3D models. This research program is expected to make substantial contributions to the solution of complex problems of high practical relevance to the field of realistic virtual world creation. It will also contribute to the development of innovative methods in the general fields of computer vision, computer graphics and computer games.