作者: Roy Henry Jakola , David Owen Parry , Shuen Yan Stephen Se , Piotr Jasiobedzki
DOI:
关键词: Projector 、 Motion (physics) 、 Matching (graph theory) 、 Point cloud 、 Track (disk drive) 、 Stereo camera 、 Artificial intelligence 、 Monocular 、 Computer vision 、 Geography 、 Mobile device 、 Computer graphics (images)
摘要: The present invention provides a system (method and apparatus) for creating photorealistic 3D models of environments and/or objects from plurality stereo images obtained mobile camera optional monocular cameras. cameras may be handheld, mounted on platform, manipulator or positioning device. automatically detects tracks features in image sequences self-references the 6 degrees freedom by matching to database track motion, while building simultaneously. A motion estimate also provided external sensors fused with computed images. Individual pairs are processed compute dense data representing scene transformed, using estimated into common reference together. resulting is represented as point clouds, surfaces, volumes. enhancing registering information additional improve model fidelity augment it supplementary light pattern projector. generating photo-realistic underground such tunnels, mines, voids caves, including automatic registration pre-existing maps.