A fast machine vision approach for automatic recognition of industrial parts

Abstract

SUMMARY The problem of recognizing arbitrary parts is of considerable interest in the field of industrial automation. While it is possible to employ carriers or pallets to prearrange the parts in fixed positions for robotic assembly tasks, a vision system which can recognize the parts in random positions, orientation as well as scaling is more flexible. In this paper, we present a fast contour-based approach to planar part recognition. The recognition results of the machine vision system can be employed for automation applications such as robotic assembly and part sorting The2-D boundary of a part is initially transformed to a 1-Dθ—scurve Where tangle as a function of arc length s along the boundary. The tangent angle is directly evaluated from an approximated cubic polynomial function. A Hough-like clustering method for similarity measures is developed. It searches for a large cluster of the same tangent angle differences between the test part and the model part. The proposed method is computationally simple and requires only a small amount of memory space.