3D Reconstruction and Navigation from Video

Acuity's video based 3D imaging technology uses projective geometry to obatain shape information from multiple still images or from video frames. By moving around in a scene or circling an object with a video camera, one can collect all the information necessary to reconstruct the scene or object. Our passive systems combine cameras and motion sensors to accurately reconstruct features and objects.

The bottleneck in the availability of ubiquitous 3D presentation and interaction materials is the acquisition of accurate 3D models from real world spaces. If such models can be rapidly and automatically created with video sequences from a freely moving video camera, this bottleneck can be removed. In recent years a unified approach has emerged for creating three dimensional models of world objects and spaces from multiple two dimensional views, such as those obtained from a video camera. This has opened the door to capture and re-creation of real world environments. Computer displays and processing power have reached the point where realistic 3D presentation with smooth motion is available to most home and business users. Increased use of broadband internet connections in the home is enabling transmission and viewing of detailed interactive graphic material.

Our hardware and algorithms provide tools for the automation of world space capture through video sequences. Additional information about the camera’s position and orientation is provided by inertial sensors and magnetometers collocated with the camera. With this information the task of identifying matching points in images from multiple viewpoints is moved from 2D image space into 3D world space. Point locations and camera parameters are estimated simultaneously in real time.