Arc Adjacency Matrix-Based Fast Ellipse Detection.

Area of Science:

• Computer Vision
• Image Processing
• Pattern Recognition

Background:

• Accurate ellipse detection is crucial for various computer vision applications.
• Existing methods often struggle with speed and precision, especially in complex image data.
• There is a need for robust and efficient algorithms for identifying elliptical shapes.

Purpose of the Study:

• To propose a novel and efficient ellipse detection method.
• To enhance the accuracy and speed of ellipse detection in computer vision tasks.
• To introduce the arc adjacency matrix-based ellipse detection (AAMED) method.

Main Methods:

• Elliptic arc segmentation and construction of a digraph-based arc adjacency matrix (AAM).
• Utilizing curvature and region constraints to sparsify the AAM.
• Bidirectional searching of the AAM for ellipse candidates and calculating cumulative matrices (CM) using cumulative factors (CF).
• Efficient ellipse fitting via eigendecomposition of CM and validation using a comprehensive score.

Main Results:

• The proposed AAMED method demonstrates superior performance over 12 state-of-the-art methods across 9 datasets.
• Achieved improvements in recall, precision, F-measure, and reduced time consumption.
• Effectively eliminated false ellipses through a proposed validation score.

Conclusions:

• The arc adjacency matrix-based ellipse detection (AAMED) method offers a significant advancement in computer vision.
• The method provides a fast, accurate, and robust solution for ellipse detection.
• AAMED outperforms existing techniques, making it suitable for demanding applications.