关键词: 蔬菜, 采摘机器人, 视觉系统, 末端采摘装置, 发展趋势

Abstract:  The vegetable picking robot plays a crucial role in the mechanization, automation, and intelligentization of vegetable production. This paper summarizes and analyzes the research status and key technologies of vegetable picking robots at home and abroad, including optimizing end effectors to improve picking efficiency and enhancing recognition accuracy through advanced image processing algorithms and deep learning models. The characteristics and suitable scenarios of three types of end effectors for vegetable picking robots, namely gripperbased, suctionbased, and biomimeticbased, are analyzed. The composition and structure of the visual system in vegetable picking robots are discussed, and three recognition methods are compared: traditional image processing methods based on color and texture features, machine learning methods such as Kmeans clustering and support vector machine (SVM) algorithms, and deep learning methods such as YOLO, Faster RCNN, and SSD networks. Based on the growth environment and characteristics of different vegetables, suitable recognition methods are summarized and compared in terms of recognition performance. Additionally, issues related to the target of operation, working environment, hardware constraints, and production costs of vegetable picking robots are identified, followed by prospective analysis on vegetable cultivation patterns, software systems, hardware systems, and leveraging regional characteristics.

Key words: vegetables, picking robot, vision system, end picking device, development trend