Detecting and Locating Partly Occluded Planar Parts by Clustering

Abstract

In this paper I address the problem of recognizing and locating partly occluded industrial parts in two-dimensional space. The shapes of parts under study consist of linear and circular segments. Most man-made objects have such geometric primitives. Many industrial parts such as stamped parts and flat foundry castings can be considered planar because of their small thickness. Therefore the restrictions on the recognized parts are widely applicable to the industrial environment.

A model-based object recognition method is introduced where the linear and circular segments are used as feature primitives to construct an attributed relational graph. Each node in the graph represents a primitive, and is described by the primitive’s geometric properties. The arc between any two nodes represents the geometric relation (distance, angle, etc.) between two primitives. In the feature extraction phase, the low-level primitives are extracted by dominant point detection, least-squares fitting and segment splitting-and-merging. In the matching phase, a Hough-like clustering procedure is developed to recognize and locate all objects in the model and scene graphs. The matching procedure does not rely on any salient features of the objects and is invariant to translation and rotation changes. The proposed method is useful for recognizing and locating severely occluded planar parts.

Additional information

Notes on contributors

Du-Ming Tsai

Du-Ming Tsai received the B.S. degree in Industrial Engineering from Tunghai University, Taiwan, R.O.C., in 1981, and the M.S. and Ph.D. degrees in Industrial Engineering from Iowa State University, Ames, Iowa, U.S.A., in 1984 and 1987, respectively. From 1988 to 1990 he was a Principal Engineer of Digital Equipment Corporation, Taiwan branch, where his work focused on process and automation research and development. He is currently Professor of Industrial Engineering at the Yuan-Ze Institute of Technology, Taiwan, R.O.C. His research interests include manufacturing automation, machine vision, and robotics.