A high-precision measurement method for ruby balls based on multi-scale adaptive edge detection

Abstract This paper proposes a high-precision measurement method for ruby balls based on multi-scale adaptive edge detection. By combining dark field imaging technology with improved Canny algorithm, the accuracy and robustness of edge extraction have been significantly improved. When measuring 5mm ruby balls, the average absolute error of this method is only 1.6µm, which is about 66% lower than the Hessian matrix method and about 83% lower than the regionprops method; When the signal-to-noise ratio drops to 20 dB, the error growth only reaches 2.4m, while the traditional Canny algorithm reaches 6.5µm. For samples with different diameters (3-12mm) and materials (ruby, ceramic, hard alloy), the algorithm's measurement error is controlled within 0.2–0.3µm, and the repeatability is better than 0.15µm, meeting the micrometer level accuracy requirements of high-end manufacturing industry. The experiment verified the excellent adaptability and robustness of this method in complex environments.