The improved YOLOv11 model based on lightweight coordinate attention module is used for automatic recognition of maize disease images

At present, agricultural production needs to be transformed to an efficient and sustainable model, in which pest identification and control are the key links to ensure crop yield and quality. The traditional manual inspection and pesticide application methods are inefficient, and are easy to cause pesticide waste and ecological pollution. Deep learning technology provides a new solution for agricultural disease identification. In this paper, an improved YOLOv11 model with Lightweight Coordinate Attention (LCA) was proposed for the automatic detection and accurate identification of three typical diseases (rust, gray spot and leaf spot) of maize leaves. A maize disease dataset containing 4120 high-quality images was constructed, and based on the improved CSPDarknet backbone network, feature fusion module and optimization loss function design, the accurate localization and classification of the diseased spot area were realized. Experimental results show that the YOLOv11-LCA model achieves excellent levels in many key indicators such as Precision, Recall and mAP, and is significantly better than YOLOv5~YOLOv11 and other existing models. This study not only provides an efficient and deployable solution for the intelligent identification of maize diseases, but also provides technical support for the intelligent upgrading of agricultural equipment such as field operation robots and intelligent spraying systems. In the future, more advanced detection algorithms will be further explored, detection accuracy and model deployment performance will be improved, and the practical application of the algorithm in smart agriculture will be promoted.