Circle Detection Based On Improved Hough Transform

The detection of circle including camber is one of classical problems in digital image processing, and has extensive application background. Since 1980s,researchers have provided many methods of circles detection. Hough transform is an important method. Classical Hough transform transform edge of image to a region of parametric space, and select the most accumulated edges. In real-life images, the shapes are often distorted from their true parametric forms due to the presence of noise, digitization error and shape variations. Therefore, after apply Hough transform, it may be difficult to find out a single peak, thus it is difficult to detect image.In this paper, it discusses the current Hough transform algorithm and existing problems of the algorithm, and makes use of circle property to provide improved algorithm based on Hough transform.This paper emphasizes on introducing random Hough transform theory,takes three-dimensional parametric of the tradition Hough transform on detection of circle to one-dimensional statics of random Hough transform and introduces fuzzy theory, to every parameter's acculate compute of parametric space,is by accurate fuzzy membership that fuzzy edge of image space.This abstains the problem only by acculating edge which the fuzzy membership equal to one.Theories and experiments show that the paper presents an improved Hough transform algorithom, which has better detection performance and can get accurated result.