A new Method to Detect Circles in Images Based on Genetic Algorithms

Abstract

Object detection is one of the key issues in digital image processing. Over the years, many algorithms have been created for detecting meaningful objects on the image which are based on specific characteristics of object or complex mathematical methods. Circle detection is one of these types of methods. One of the best methods for circle detection on digital images and discussion of machine vision is the Hough transform. The Hough transform can be described as the transformation of a point in the x-y-plane to the parameter space. Parameter space can be defined by the shape of the object. Using the special character of each image in space, we are able to retrieve and extract the image circle. Importantly, this method is time consuming and a large amount of memory is required for the image. The undesirable features have reduced the popularity of this method. The idea of using genetic algorithm for detecting a circle in the picture is very attractive and functional. This method can be used in Robot Soccer, targeting systems and iris recognition. In this method, accuracy and speed are among important parameters. For example, in the case of robot, it should detect ball in monochromatic and sometimes crowded areas (due to accumulation of other bots around the ball). Using a genetic algorithm for circle detection on images, Hough transform weaknesses have been removed. It also increases the computation speed and accurate detection of circle. In this paper, the Hough transform method will be presented and then we will describe the process of implementing genetic algorithms to find a circle in the picture.