Abstract—This paper is devoted to the analysis and implementation of the algorithms for automatic detection of the circular objects in the image. The practical aim of this task is development of the algorithm for automatic detection of log abuts in the images of roundwood batches. Based on literature review four methods were chosen for the further analysis and the best performance out of them was provided by ELSD algorithm. Some modifications were implemented to the algorithm to fulfill the requirements of the given task. After all, the modified ELSD algorithm was tested on the dataset of the images. The relative accuracy of the algorithm in comparison with manual measurement is 95.2% for the images with total area of background scene less than 20%.
Index Terms—Arcs grouping, DBSCAN, ELSD algorithm, equivalent circle, hough transform.
A. V. Kruglov is with the Ural Federal University, Yekaterinburg, Russia, 620004 (e-mail: firstname.lastname@example.org).
Cite: Artem V. Kruglov, "The Unsupervised Learning Algorithm for Detecting Ellipsoid Objects," International Journal of Machine Learning and Computing vol. 9, no. 3, pp. 255-260, 2019.