An online automated grading system of pearl perfection based on computer vision was introduced to replace time-consuming human grading. Firstly, the whole pearl surface was captured as images when the pearl freely fell. The surface images were then analyzed in real-time to obtain valid features, and then the features were applied to determine the final pearl perfection grade. The performance shows the new system greatly improves the efficiency of pearl surface perfection grading and cuts the cost down as well.