Development and Field Evaluation of an Autonomous Runner Cutter System Prototype for Strawberry Production in Plasticulture

Authors:
Preprint
Publisher website
Google Scholar
Abstract
Excessive runners (stolon) in strawberry plants divert resources from mother plants and they can absorb soil-borne pathogens and herbicide residues from the unfumigated row-middle soil, all of which threaten yield and plant health. Manual runner removal is labor-intensive and time-consuming, causing additional costs per acre for growers. The goal of this project is to develop and field-evaluate a prototype autonomous strawberry runner cutting system by integrating computer vision, decision making, advanced robotics, and vehicular platform to automate the runner cutting process to increase its efficiency and reduce the labor demand. For runner detection, a YOLO-based model was trained using 3,036 high-resolution RGB runner images. For runner 3D localization, 3D point cloud-based distance estimation technique was performed. The cutting mechanism was implemented by a six DOF robotic arm with cutting blades. These modules were then integrated into an autonomous field vehicle. The utilization of the state-of-the-art sensor and computer vision algorithms ensured the precise 3D localization of the runners for cutting, while leaving the necessary components of the plant unharmed. The detection module was tested with three strawberry cultivars: Brilliance, Ember, and Encore; and achieved the highest precision of 86.7%, recall of 49%, and F1 score of 62.7%. The runner localization module estimated the distance with high precision compared to the ground truths (R2 = 0.9147 and error median 2 mm). Cutting efficiency of only the cutter was 83.87%, and the overall system efficiency was 48.15% for all cultivars combined. Average data pre-processing time of 12.05 seconds, average runner inference time of 2.20 seconds, and average cutting time per runner of 18.70 seconds were recorded for the system. After successful completion of this study, it is expected to have a field-deployable autonomous runner cutter prototype with great potential to support healthier plants, increased fruit yields, and reduced manual labor.