Chessboard Corner Detection Based on EDLines Algorithm.

Area of Science:

• Computer Vision
• Robotics
• Image Processing

Background:

• Accurate camera calibration is crucial for various applications, including robotics and computer vision.
• Traditional chessboard corner detection methods can suffer from robustness and accuracy issues, especially in complex imaging conditions.

Purpose of the Study:

• To propose a novel camera calibration method for robust and accurate automatic detection of chessboard corners.
• To leverage the EDLines algorithm for enhanced straight-line detection and corner identification.

Main Methods:

• Utilizing the EDLines algorithm for initial straight-line detection in calibration images.
• Filtering detected lines based on corner features to exclude background elements.
• Employing gray gradient sorting and line fitting for initial corner coordinate estimation.
• Performing subpixel optimization and pixel-coordinate system conversion for precise corner localization.

Main Results:

• The proposed method demonstrates no missed detections or redundancy in corner detection.
• Experiments with varying camera exposure times and complex backgrounds validate the algorithm's performance.
• Achieved an average reprojection error of less than 0.05 pixels.

Conclusions:

• The EDLines-based camera calibration method offers significant improvements in robustness and accuracy.
• The algorithm is suitable for practical applications requiring precise camera calibration.
• The refined corner detection technique contributes to more reliable machine vision systems.