A machine vision navigation based autonomous weeding robot in the agricultural fields is developed. Fuzzy control is employed to guide the robot to move along the crop line. Membership functions are chosen according to the properties of navigation angle and navigation distance. Two control rule banks are constructed and their control results are discussed. Experiments show that the fuzzy control could make the robot move steadily. The rule 1 makes the robot move accurately along a straight line. The rule 2 makes the robot move smoothly along a curve line; when the robot moves along a 42.2° curve line, the accuracy is 74.58%.