Aimed at the low automation level of tomato picking, a tomato picking system combining binocular vision acquisition system and robot positioning technology was designed.Binocular stereo vision system was used for image acquisition, model construction and calibration.The image was preprocessed by image processing technology, so as to obtain the specific contour of the mature tomato.According to the tomato ellipse shape, the coordinates of the center point were determined and converted into space coordinates. It was convenient for the robot end effector gripper to locate and pick. Aimed at the motion limitation of each joint, the smoothness constraint was proposed. The ant colony algorithm was improved, and a smooth and shortest path for the picking gripper was obtained. The simulation results show that the method can basically meet the requirements of picking image processing,and can obtain the shortest and smooth path.It provides reference for realizing tomato fully automatic picking