Addressing the inefficiencies and inaccuracies inherent in manual measurements of flange disk dimensions, we propose a machine vision-based method for sub-pixel measurement, alongside a dedicated visual measurement system. This system utilizes a CCD camera, calibrated via PyCharm software, to capture images of the flange disk. These images undergo preprocessing—conversion to grayscale, filtering, noise reduction, and binary segmentation—to delineate the flange disk area for further analysis. For edge detection, an improved Canny adaptive threshold selection algorithm is employed. Additionally, we introduce a new edge judgment criterion based on the Zernike algorithm, which, in conjunction with the advanced Canny operator for initial positioning, allows our refined Zernike matrix algorithm to accurately identify the flange disk edge with sub-pixel precision. The least squares fitting algorithm is then applied to determine the flange disk's outer and inner diameters accurately. A comparative analysis of experimental results demonstrates that our enhanced algorithm significantly reduces relative errors compared to both its predecessor and manual measurements. Importantly, the measurements obtained with the improved algorithm closely match manual measurements and are markedly more accurate than those achieved with the traditional Zernike matrix algorithm. This level of precision ensures that our measurement method meets the high accuracy requirements of industrial parts manufacturing.