A LIGHTWEIGHT WEED IDENTIFICATION METHOD FOR CHINESE CABBAGE FIELDS BASED ON THE IMPROVED YOLOv11

Accurate and real-time weed identification was a key technology for ensuring the efficient operation of field weeding robots. It played a vital role in reducing pesticide usage, minimizing environmental pollution, and protecting the agricultural ecosystem. In response to these demands, this paper proposed a lightweight and high-precision weed recognition method, YOLO-LMSW, tailored for Chinese cabbage field scenarios. Firstly, a lightweight multi-scale convolutional module (LMSConv) was designed, upon which the LMSC3k2 structure was constructed to reduce computational complexity and enhance multi-scale feature extraction capabilities. In terms of network architecture, a lightweight backbone network, LMSNet, was built to significantly reduce model parameters while maintaining detection accuracy. Additionally, a detection head, LMSHead, was designed to further optimize the model structure. To improve localization accuracy and convergence speed, the Wise-IoU (WIoU) loss function was introduced. Experimental results demonstrated that, compared to YOLOv11n, YOLO-LMSW achieved improvements of 1.2%, 1.0%, and 0.6% in precision, recall, and mAP50 respectively, while reducing the Params and FLOPs by 34.6% and 36.5%, respectively. Furthermore, the model was deployed on a Jetson Orin Nano (4GB) embedded device, where it achieved an average inference time of 117 ms per image. These results demonstrated its application potential in the actual deployment of field weeding robots.