Computer vision-based high-quality tea automatic plucking robot using Delta parallel manipulator

• Published in: Computers and electronics in agriculture 2021-02, Vol.181, p.105946, Article 105946
• Main Authors: Yang, Hualin, Chen, Long, Ma, Zhibin, Chen, Miaoting, Zhong, Yan, Deng, Fang, Li, Maozhen
• Format: Article
• Language: English
• Subjects: Algorithms; Ant colony optimization; Automatic control; Computer vision; Delta parallel manipulator; Machine learning; Manipulators; Motion control; Particle swarm optimization; Path planning; Plucking; Recognition; Robot arms; Support vector machines; Tea plucking robot; Trajectory planning
• ISSN: 0168-1699; 1872-7107
• DOI: 10.1016/j.compag.2020.105946

•A complete solution of the high-quality tea automatic plucking robot is presented.•The PSO-SVM algorithm is proposed to recognize the tender tea shoots.•The YOLOv3 model is developed to localize the plucking point.•The improved ACA is developed for path planning of the Delta parallel manipulator.•The feasibility and effectiveness of the presented solution is studied. Aiming to reduce the labour cost and improve the competitiveness of the tea production, this paper presents a complete solution, including the mechanical structure, the visual recognition system (VRS) and the motion control system of the high-quality tea automatic plucking robot. The design objectives involve high reliability and stability, ease of maintenance, cost-effectiveness, accurate recognition and localization of tea shoots, and high-efficient plucking manipulation. The VRS is established using the computer vision together with corresponding machine learning algorithms, where the Particle swarm optimization-Support vector machine algorithm and the deep convolutional neural YOLOv3 model are developed to recognize the tender tea shoots and localize the plucking point, according to the images captured by the global and local cameras, respectively. The Delta parallel manipulator equipped with a shear cutter is adopted to execute the trajectory tracking and tea plucking operations. The improved ant colony algorithm is utilized for path planning and trajectory smoothing of the manipulator. Simulation and laboratory platform experiment results verify the feasibility and effectiveness of the presented solution.