Object recognition and localization via pose clustering

摘要: The general paradigm of pose clustering is discussed and compared to other techniques applicable the problem object detection. Pose also called hypothesis accumulation generalized Hough transform characterized by a “parallel” low level evidence followed maxima or step which selects hypotheses with strong support from set evidence. Examples are given showing use in both 2D 3D problems. Experiments show that positional accuracy points placed data space model obtained via comparable sensed candidates computed. A specific sensing system described yields an few millimeters. Complexity approach relative alternative approaches reference conventional computers massively parallel computers. It conjectured can produce superior results real time on machine.