Research in Information Science & Technology: Machine Vision

摘要

Machine Vision is fast becoming a key technology, and advances in machine vision are occurring along several fronts. This report outlines the progress being made at Columbia University in developing new machine vision algorithms and applications, and the associated technology transfer. Specific results include a new model of Lambertain reflectance, methods for recovery of shape from specularity, integrating color and polarization for shape recovery, visual learning of appearance for fast object recognition, automated 3-D model acquisition from range imagery, new methods for modeling deformable objects, deriving shape from shadow information, methods to control robotic hands with vision, new approaches to sensor planning and placement, generating spatial language descriptions from imagery, and vision algorithms to recognize hand gestures.