An Intelligent Robot Based on Optimized YOLOv11l for Weed Control in Lettuce

Abstract

<b><sc>Abstract.</sc></b> Lettuce is a key dietary vegetable, and efficient intra-row weed control is vital for optimizing yield. However, rising labor costs, environmental hazards of herbicides, and the expense of laser-based systems hinder automated solutions. This study presents an intelligent intra-row weeding robot integrating an optimized YOLOv11l-based vision system and a mechanical weeding platform. The model achieved high detection metrics (Precision 99.699%, Recall 99.863%, mAP50 99.499%), outperforming other YOLOv11 variants. In conveyor belt tests, the system attained 84.579% lettuce localization and 81.058% weed removal rates across variable weed densities and speeds. Field trials confirmed robust performance, achieving up to 82.218% weeding efficiency and injury rates below 2.6% under varying lighting and travel conditions. The proposed system offers a high-accuracy, efficient solution for precision lettuce weeding, promoting sustainable agricultural practices.