As a kind of intelligent agricultural equipment, picking robots are of great significance for improving the efficiency of agricultural production. However, the main bottleneck restricting the development of picking robots today is the positioning and control in image recognition. Therefore, an agricultural picking control method based on visual servo technology is proposed. This method can accurately control the picking hand of the picking robot on the basis of building an eye-hand relationship model and an online identification system. With the tomato picking process as the background, the effectiveness of the method was verified. The test results of image feature points show that there is a small error between the stable feature points and the expected image feature points. In addition, the image plane trajectory of the picking robot is relatively smooth, and there is no vibration or overshoot. In addition, from its density distribution characteristics, it can be seen that the picking hand movement is a continuous acceleration stage at the beginning of the control, and at the end of the control, the characteristic points gradually tend to the desired characteristics, and the output of the control system is relatively stable. It can be seen from this that the model has better control performance. This method has certain practical significance for the improvement of agricultural picking efficiency.