农业机械学报2026，Vol.57Issue(9)：246-253,8.DOI:10.6041/j.issn.1000-1298.2026.09.023
连续两段式果蔬采摘机械臂视觉伺服控制方法研究
Research on Continuous Two-stage Visual Servo Control for Robotic Arm in Fruit and Vegetable Harvesting
摘要
Abstract
In the apple production process,the labor cost incurred during picking typically represents approximately 30％to 50％of the overall fruit harvest expenditure for the entire orchard.According to statistics,the mechanization rate of apple picking operations in China is less than 3％,most of them are still picked manually,the picking operation space is small,the danger is high in the working process,and the picking efficiency is low.It can be seen that the low mechanization of orchards restricts the development of orchard industry in China.However,the current manipulator picking mode of"observing before picking"has difficulty in adapting to the unstructured environment of orchards and external disturbances.In order to improve the picking speed and positioning accuracy of picking manipulator in unstructured environments,visual servo technology has been widely applied in picking operations.However,the current monocular visual servo control methods have problems such as low positioning accuracy and poor robustness.A continuous two-stage visual servo control method based on images and combined a monocular camera and a laser distance sensor was adopted to achieve efficient picking of red apples.The research gathered a dataset comprising 1 800 apple images captured at diverse times throughout the day and under assorted weather circumstances.Each image encompassed multiple apples and exhibited features such as uneven natural lighting,overlapping,and occlusion by branches and leaves.Data augmentation techniques were applied to employ the dataset to four times its initial volume,enhancing the generalization ability and recognition accuracy of the training model.The detection phase employed the YOLO v5s network to detect the target fruit,and then application of threshold segmentation to extract the target area.In the visual servo process,the center point of the target fruit served as the image feature,and the speed control rate of the robotic arm joint was formulated by combining PD control.The visual servo process was segmented into alignment stage and approach stage,and the target depth information in the two stages was obtained from the radius estimation of the target in the image and the measurement data of the laser ranging sensor,respectively.In order to ensure the continuity of the longitudinal velocity during the phase switching,the dynamic weight method was adopted to smooth the velocity in the transition zone.In order to ensure the stability of the manipulator when approaching the singular position region during the picking process,the joint velocity was calculated by the dynamic damping least squares method.In the picking experiment conducted along the non-singular trajectory,the maximum absolute positioning error in the absence of disturbances was 4.00 mm,and under the condition of external disturbances,the maximum absolute positioning error reached 5.27 mm.This continuous two-stage control approach,in contrast to the traditional segmented visual servo,the proposed continuous two-stage control method managed to reduce the visual servo control time of the robotic arm by 43.36％,in the singular trajectory picking experiment,the maximum absolute positioning error was merely 1.53 mm,and the roll-pitch-yaw(RPY)value was altered by(0.085 9,-0.034 5,-0.105 2)rad,this was due to the fact that the dynamic damping least square method sacrificed the end pose accuracy to ensure the stability of the joint speed of the manipulator in the singular configuration region.The experimental outcomes indicated that the proposed visual servo control method substantially enhanced the control speed and robustness of visual servo.关键词
果蔬采摘机械臂/视觉伺服/深度估计/速度平滑/奇异位形Key words
robotic fruit and vegetable harvesting/visual servo/dynamic weighting/velocity smoothing/singularity分类
农业科技引用本文复制引用
贺磊盈,李洁,蒋林祥,杜小强,李亚涛..连续两段式果蔬采摘机械臂视觉伺服控制方法研究[J].农业机械学报,2026,57(9):246-253,8.基金项目
浙江省自然科学基金重大项目(LD24E050006)和浙江省"领雁"研发攻关计划项目(2022C02057) （LD24E050006）
