AN APPROACH FOR GENERIC DETECTION OF CONIC FORM

In this paper we introduce a unique methodology to detect any conic form by using Hough transform which is an established method for shape detection. This technique works based on features extraction for each conic separately and intend to unify all type of Hough detection in only one. Once we have the conic’s parameters it is possible to build an accumulator array named parameter space where the most voted set of parameters is the chosen form. To build this accumulator, a range of values is set up and each value is a real parameter candidate. The length of which cell depends on the accuracy desirable and the dimension depends on the type of conic. This transform maps points from digital space to parameters space in order to match the point of the chosen form to the form in the real image. In this way it is much easier to reconstruct the conic form whenever it is necessary and to detect a form from artificial images or handmade drawing. Although each form has one different equation and number of parameters, we realized it follows a pattern and concluded that is possible to elaborate a generic equation and a methodology to detect any conic. The use of polar coordinates makes the generic detection easier because of the rotations variety detection. In addition, the order of search is important. We observed that the detection result is better if the searching for the correct rotation is the first step. Another relevant thing is the chosen length of bins in the accumulator which really defines a better match. This paper involves analysis about each conic equation, their similarities, number of parameters and how they can be introduced by a unique expression for Hough transform.