Robotic Grapevine Winter Pruning through Vision and AI

Grapevine winter pruning is a labor-intensive and repetitive process that significantly influences the quality and quantity of the grape harvest and produced wine. Because of its complexity and repetitive nature, the task demands skilled labor that needs to be trained, as in many other agricultural sectors. This thesis encompasses two developed approaches that target using a robotic system to perform grapevine winter pruning, using a vision system and artificial intelligence. The initial phase focuses on the 2D approach, harnessing the power of 2D rgbd images with segmentation neural networks. These networks enable the identification and classification of vital plant structures. The thesis details the post-processing steps applied to the 2D results. These procedures encompass mask merging to simplify outputs alongside error detection mechanisms to eliminate small, erroneous segments. The second phase of the thesis ventures into the third dimension, introducing point clouds. In the 3D approach, depth images and segmentation masks obtained in the 2D approach converge to create instance-segmented point clouds. The study does not only delineate the integration of 2D and 3D methods but also scrutinizes their efficacy in pruning point identification. The practical implications are evaluated through real-world performance analyses. The real-world performance was analyzed via the utilization of the created system in two different pruning seasons, one during the winter of 2021/2022 and the other in the winter of 2022/2023, where the system was used in a simulated vineyard to prune a set of test plants. Moreover, the thesis underscores a unique facet of adaptability, presenting a customizable framework that empowers end-users to fine-tune parameters according to their specific requisites. This adaptability extends to variables such as the number of nodes to retain on pruned canes and the preferred cane thickness, encapsulating the versatility of the 3D approach.