Multiple straight line detection based on labeling of pixels by Genetic Algorithm
Takashi Shimamura, Manabu Hashimoto, Takayuki Fujiwara, Takuma Funahashi, Hiroyasu Koshimizu · 2011
This paper proposes a method for detecting two or more straight lines from an edge image in a similar way of Hough transform. The method consists of labeling of each edge point to presumed straight lines and RANSAC algorithm to evaluate each label. The Genetic Algorithm (GA), an efficient combinatorial optimization algorithm, is used for the labeling process, and the RANSAC is effective for high accuracy and speed-up. Some practical performances of this method have been proved through a simulation line picture with random noises.