Aiming at the problem of low grading precision of traditional automatic grading machine of distinguished and high-quality fresh tea leaves, sorting fresh tea leaves by machine vision is an effective way to improve the quality of tea grading. And a PSO-Fuzzy-PID controller introducing adaptive weights and Circle chaotic mapping was designed; the research on the control and test of conveying speed of fresh tea leaves based on improved Fuzzy-PID was carried out to address the problem of large fluctuations in the conveying speed of fresh tea leaves leading to the image blur. The results showed that the thesis algorithm had better optimization performance and convergence speed; during the operation of the conveying transmitting system of fresh tea leaves, when the conveying speed was set at 78.5mm/s, its speed was recorded every 1ms, the transmission speed fluctuation was controlled at 0.7mm/s; the transmission system response time of the improved Fuzzy-PID was reduced by 81.41% and 61.74% compared with that of the traditional PID and Fuzzy-PID, respectively; the overtone was reduced by 81.24% and 41.82%. The core contribution was to significantly improve the stability of the transport speed of fresh tea leaves, thus effectively reducing the degree of image blur caused by speed fluctuations and making the acquired images closer to the clear state. Therefore, the research result not only ensured that the automatic bud and leaf classification system based on machine vision can be accurately and stably controlled, but also provided a practical technical solution to solve the scientific problem of image blurring caused by the fluctuation of conveying speed.