Research On The Vision Control And Trajectory Planning Of Fruit Picking Robot

Agricultural robot is one of the important equipments for precision agriculture in the 21 st century and becomes a hot direction of development of intelligent agriculture machine in the future. Based on full understanding of the research and application status of fruit picking robot, an autonomic picking system was designed, which focused on recognition and positioning of fruits, hand-eye calibration, algorithm of visual control and trajectory planning and so on. Main research contents and conclusions are following:1. The fruit objects were recognized in image space by using the Hough transform. Some pixels of each object obeying the uniform distribution in the image space were sampled randomly. And then the sphere model of fruit was constructed in 3D space with these pixels, the coordinates of the center of each fruit object were obtained directly. Experimental results showed the positioning accuracy of the fruit object was better than 8mm, the robustness of algorithm to the partial occlusion and invalidation of stereo matching in the presence of complex scene.2. A convenient and effective hand-eye calibration method was developed. The transform matrix of eye and robot coordinate was acquired by least square method. Experimental results showed the calibration errors of each coordinate vector was less than 10mm. This calibration algorithm was easy to operate without expensive laboratory equipment, which was suitable for on-site calibration.3. An idea of combining open-loop and close-loop was introduced and a multi-camera visual servoing structure was built, which reduced the impact of system accuracy caused by the time delay and the estimation of depth for image Jacobian. Experimental results show that the image error converged within 5 pixels.4. The accuracy monocular measurement of distance was discussed. The measurt error was within±30mm when the distance ranges from 300mm to 600mm, which provided a method to estimate the depth for acquiring the image Jacobian matrix.5. The problem for multi-fruit picking order was studied and the best sequence was achieved using genetic algorithm, which improved efficiency and reduced power consuming. The use of cubic polynomial interpolation was explored for the trajectory planning over mid-point and the real-time trajectory planning between the two fruits, under the control of visual feedback, was achieved using the heuristics algorithm to choose the speed of mid-point automatically based on the detailed analysis of the kinematics of the fruit picking robot. The experimental results showed that the velocity of joints was continuous at any time.This thesis provides viable way for the design of fruit picking robot, and the automatic picking system meets the need of the actual picking in the reliability and precision of fruit gripping.