A METHOD FOR 3-D OBJECT POSE FROM SINGLE IMAGES IS PRESENTED. A VIRTUAL VIEWPOINT ANALYSIS SHOWS THAT A FEATURE-CENTERED COORDINATE SYSTEM DRASTICALLY REDUCES THE MATHEMATICAL COMPLEXITY OF THE INVERSE PROJECTION PROBLEM FOR ELLIPSES. PARAMETRIC MATCHING BY GENERALIZED FEATURES SOLVES FOR VIEWPOINTS. THIS PAPER REVIEWS: SOLVING THE PROBLEM OF ELLIPSE-TO-ELLIPSE CORRESPONDENCE, THUS GREATLY EXTENDING THE CLASS OF RECOGNIZABLE OBJECTS; GENERALIZED FEATURES WHICH CAN ASSUME VARIOUS CONFIGURATIONS, THUS ENABLING THE SYSTEM TO DETECT A MULTITUDE OF FEATURES; PARAMETRIC MATCHING INSTEAD OF POINT-POINT MATCHING WHICH MAKES THE SYSTEM ROBUST AND THE TOP-DOWN-BOTTOM-UP STRATEGY WHICH COMBINES EFFICIENCY WITH ROBUSTNESS.
| Number of Pages : | 24 |
| Published : | 06/01/1990 |
