In order to improve the accuracy and immediacy of particle size and distribution detection during the preparation of fresh pellets, a fresh pellets size recognition system is developed through computer vision technology. The system acquires real-time images of the ball outlet area of the disc ball maker through a camera, uses quadruple enhancement and Gaussian filtering for image enhancement and noise reduction, and then uses the Hough circle detection algorithm for fresh pellets number and particle size identification. The results show that the quadruple image enhancement combined with Gaussian filtering can effectively reduce the noise in the image, and the Hough circle detection can quickly and accurately identify the fresh pellets number and particle size distribution, and the accuracy of fresh pellets number and qualified fresh pellets number identification reaches 96.29% and 90.61%, respectively. The research results of this paper have a positive effect on improving the qualification rate of fresh pellets and reducing the labor intensity of personnel.