Objective In order to avoid seedling and reduce seedling damage rate during the operation of mechanical weeding, a control system of avoiding seedling based on machine vision and proportion integration differentiation (PID) control technology was designed. Method The improved extra-green algorithm was used to gray rice seedlings. The image projection method was used to extract the characteristics of rice seedlings within the region of interest (ROI) to obtain the corresponding image coordinates. The robust regression algorithm was used to fit rice seedlings to obtain the center line of seedling belt. The ground coordinate position of seedling, and the distance between the center of weeding unit and the center line of seedling belt were obtained by transforming the model of aperture imaging. The hydraulic rectifying system was controlled based on PID control algorithm, and the Simulink in Matlab software was used for the simulation analysis of the system. Result The seedling avoidance was realized, and the steady-state response time of the system model was 0.34 s. The performance comparison tests of the control system with and without the control system of avoiding seedlings showed that the control system of avoiding seedlings could obviously reduce the seedling damage of weeding components, with the average seedling injury rate of 3.75%, and the rate without the control system of avoiding seedlings was 24.88%. Conclusion The row-follow control system for mechanical weeding designed in this study can correct the working path of the weeding components in real time, which effectively reduces the seedling injury rate of mechanical weeding. The results of this study can provide some reference for mechanical weeding row-follow control of rice and other crops.