Edge and corner identification for tracking the line of sight

Abstract

This article presents an edge-corner detector, implemented in the realm of the GEIST project (an Computer Aided Touristic Information System) to extract the information of straight edges and their intersections (image corners) from camera-captured (real world) and computer-generated images (from the database of Historical Monuments, using observer position and orientation data) -- Camera and computer-generated images are processed for reduction of detail, skeletonization and corner-edge detection -- The corners surviving the detection and skeletonization process from both images are treated as landmarks and fed to a matching algorithm, which estimates the sampling errors which usually contaminate GPS and pose tracking data (fed to the computer-image generatator) -- In this manner, a closed loop control is implemented, by which the system converges to exact determination of position and orientation of an observer traversing a historical scenario (in this case the city of Heidelberg) -- With this exact position and orientation, in the GEIST project other modules are able to project history tales on the view field of the observer, which have the exact intended scenario (the real image seen by the observer) -- In this way, the tourist “sees” tales developing in actual, material historical sites of the city -- To achieve these goals this article presents the modification and articulation of algorithms such as the Canny Edge Detector, SUSAN Corner Detector, 1-D and 2-D filters, etcetera