Pyramidal Blur Aware X-Corner Chessboard Detector

With camera resolution ever increasing and the need to rapidly recalibrate robotic platforms in less than ideal environments, there is a need for faster and more robust chessboard fiducial marker detectors. A new chessboard detector is proposed that is specifically designed for: high resolution images, focus/motion blur, harsh lighting conditions, and background clutter. This is accomplished using a new x-corner detector, where for the first time blur is estimated and used in a novel way to enhance corner localization, edge validation, and connectivity. Performance is measured and compared against other libraries using a diverse set of images created by combining multiple third party datasets and including new specially crafted scenarios designed to stress the state-of-the-art. The proposed detector has the best F1- Score of 0.97, runs 1.9x faster than next fastest, and is a top performer for corner accuracy, while being the only detector to have consistent good performance in all scenarios.