Abstract
Shape recognition is an important task in the field of computer vision. Many algorithms were developed to extract features from captured images of valuable objects. The traditional polygonal approximation algorithm (PAA) is a well-known algorithm in shape analysis for digital image processing and pattern recognition. The main objective of this algorithm is to convert a set of connected points in digitized images into a list of fitted lines, which represent a two-dimensional polygon. The main problem with this algorithm is that it lacks the ability to extract curved objects from the digitized images. In this paper, a new algorithm called an enhanced polygonal approximation algorithm (EPAA) is developed, based on the PAA, to extract both straight and curved features from digitized images of two-dimensional products. The EPAA is capable of classifying line and arc segments as well as circles. A description of the EPAA and an example of its application in the field of production engineering are presented.