Showing papers on "3D single-object recognition published in 1980"

Computer Architecture for Object Recognition and Sensing

Abstract: The notion of using many, most likely different, sensory subsystems in a computer object recognition system immediately provokes several questions: How will multiple sensors be used in conjunction? What object qualities are best described by which sensor, and how is sensor utilization optimized? To what extent does the information provided by each sensor overlap with that provided by others, and how then is this used? Comments University of Pennsylvania Department of Computer and Information Science Technical Report No. MSCIS-80-22. This technical report is available at ScholarlyCommons: http://repository.upenn.edu/cis_reports/724 University of Pennsylvania The Moore School of E lec t r ica l Engineering Computer Architecture f o r Object Recognition and Sensing

Three-Dimensional Object Recognition System: Ranging Camera And Algorithm

Abstract: A scanning triangulation -based laser range- finder is described, along with a preliminaryversion of an algorithm for 3 -D object recognition The ranging camera employs a novel de-sign which should ultimately enable a range sample rate of approximately 100 kHz with anaccuracy of lcm at ranges <3m The system employs a spherical coordinate scanning geometry;a laser beam emerges vertically from a hollow shaft and is reflected by a cubic mirrorrotating on a horizontal shaft, yielding a vertical (polar angle) scan A motor rotates theentire system through a slower 360° azimuthal scan A detection system including a slittedwheel and photo- multiplier, located meter above the mirror, repeatedly measures the anglebetween the vertical and the line of sight to the laser spot This and the vertical scanangle is used by an LSI -11 computer to compute range The field of view is a broad "equa-torial" band of nearly 37 steradians An algorithm is presented for recognition of objectsknown to the system The surface of each object is approximated by a union of convexpolyhedra, represented as a Boolean combination of linear inequalities A shell is producedenclosing the surface but not the interior Then sets of contiguous points from a range -picture are tested for consistency with some rotation and translation of the polyhedralmodel of each object The algorithm is tolerant of occlusion and random errorsIntroductionThe automatic recognition of solid objects in a scene is a difficult problem if only con-ventional brightness images are used The extraction of 3 -D geometric information from suchpictures requires knowledge of the physics of reflection as well as complicated word -know-ledge about probable patterns of scene illumination and surface reflectance characteristicsOn the other hand, range -pictures directly provide 3 -D coordinates of surface points in a