We present what is to our knowledge the first-ever fitting-based circle detection algorithm, namely, the fast and accurate circle (FACILE) detection algorithm, based on gradient-direction-based edge clustering and direct least square fitting. Edges are segmented into sections based on gradient directions, and each section is validated separately; valid arcs are then fitted and further merged to extract more accurate circle information. We implemented the algorithm with the C++ language and compared it with four other algorithms. Testing on simulated data showed FACILE was far superior to the randomized Hough transform, standard Hough transform, and fast circle detection using gradient pair vectors with regard to processing speed and detection reliability. Testing on publicly available standard datasets showed FACILE outperformed robust and precise circular detection, a state-of-art arc detection method, by 35% with regard to recognition rate and is also a significant improvement over the latter in processing speed.
© 2013 Optical Society of AmericaFull Article | PDF Article
Christopher A. Cook and Jane F. Koretz
J. Opt. Soc. Am. A 15(6) 1473-1485 (1998)
Haogang Zhu, David P. Crabb, Patricio G. Schlottmann, Tuan Ho, and David F. Garway-Heath
Opt. Express 18(24) 24595-24610 (2010)
Sribalamurugan Sekhar, Fathi E. Abd El-Samie, Pan Yu, Waleed Al-Nuaimy, and Asoke K. Nandi
Appl. Opt. 50(19) 3064-3075 (2011)