OBJECT RECOGNITION

In a method and apparatus for the location and recognition of a simple (e.g. polyhedral) object (2), relative movement between the object and a video camera (1) is achieved by scanning of the camera or movement of the object. The object is surveyed by the camera, and pixels representing at least thr...

Full description

Saved in:
Bibliographic Details
Main Authors DAVID WILLIAM MURRAY, DAVID ARTHUR CASTELOW, BERNARD FRANCIS BUXTON
Format Patent
LanguageEnglish
Published 07.06.1989
Subjects
CALCULATING
COMPUTING
COUNTING
PHYSICS
Online AccessGet full text

Cover

More Information
Summary:In a method and apparatus for the location and recognition of a simple (e.g. polyhedral) object (2), relative movement between the object and a video camera (1) is achieved by scanning of the camera or movement of the object. The object is surveyed by the camera, and pixels representing at least three successive images (3) generated by the camera are stored. The pixels are processed to locate edge-representing pixels (edgels). Visual motion vectors are computed at the edgels in the central one of the three images. The edgels are notionally linked into strings (18-36) representing edges of the object interconnected at vertices (12-17), to form a vertex-edge graph, which is subsequently segmented into topologically connected subgraphs. For each subgraph the distances of the vertices from the camera and their velocities relative to the camera are computed, and a partial wireframe of the object is derived from those distances. The partial wireframe is then compared with stored data relating to wireframe models to determine whether there is correspondence between the shape of the object and the shape represented by any of the models.
Bibliography:Application Number: GB19870021425