3D INFORMATION EXTRACTION FROM VIDEO SEQUENCES THROUGH A SINGLE CAMERA

摘要

PURPOSE: A 3D space information extraction system from an input video image using a single camera is provided to extract a 3D space from a consecutive group of 2D images. CONSTITUTION: A feature point is recognized using an extended Kalman filter algorithm. A location on a 3D area of the recognized feature point is extracted. Before and after frames of an input inputted through one camera are compared. The feature point commonly included in each frame is extracted. A 3D space is restored by only image information through projective geometry.