Robust Detection of Non-overlapping Ellipses from Points with Applications to Circular Target Extraction in Images and Cylinder Detection in Point Clouds

TL;DR

This paper introduces new robust methods for detecting non-overlapping ellipses in 2D edge points and extracting cylinders from 3D point clouds, demonstrating superior accuracy and robustness over existing techniques.

Contribution

The paper presents novel Monte Carlo-based ellipse fitting and detection methods, along with a new approach for cylinder extraction from 3D point clouds, validated through extensive experiments.

Findings

Robust ellipse detection outperforms four existing methods in accuracy.

Achieved 99.3% F-measure in real image ellipse detection.

Successfully identified all detectable pipes in real-world point clouds.

Abstract

This manuscript provides a collection of new methods for the automated detection of non-overlapping ellipses from edge points. The methods introduce new developments in: (i) robust Monte Carlo-based ellipse fitting to 2-dimensional (2D) points in the presence of outliers; (ii) detection of non-overlapping ellipse from 2D edge points; and (iii) extraction of cylinder from 3D point clouds. The proposed methods were thoroughly compared with established state-of-the-art methods, using simulated and real-world datasets, through the design of four sets of original experiments. It was found that the proposed robust ellipse detection was superior to four reliable robust methods, including the popular least median of squares, in both simulated and real-world datasets. The proposed process for detecting non-overlapping ellipses achieved F-measure of 99.3% on real images, compared to F-measures of 42.4%, 65.6%, and 59.2%, obtained using the methods of Fornaciari, Patraucean, and Panagiotakis, respectively. The proposed cylinder extraction method identified all detectable mechanical pipes in two real-world point clouds, obtained under laboratory, and industrial construction site conditions. The results of this investigation show promise for the application of the proposed methods for automatic extraction of circular targets from images and pipes from point clouds.