Authors

Abstract

Accurately counting the number of corn seedlings is the key to evaluating the growth status of corn. To address the problem of difficult detection and counting of corn seedlings in complex field environments, this study proposes an improved MEI-YOLOv11 model. By introducing MANet, EUCB module, and Inner-SIOU loss function, the ability to extract features and recognize small targets in complex environments is enhanced. The results showed that the mAP0.5, P, and R of the model reached 97.0%, 94.2%, and 95.7%, respectively, which were 2.8, 2.7, and 2.4 percentage points higher than YOLOv11, respectively. The parameter count and inference time only increased by 1.28 M and 0.4 ms, respectively, and the detection accuracy was better than other detection models. The accuracy of multi weather counting is above 90%, with the highest accuracy of 91.23% on sunny days (RMSE=4.5044, R ²=0.8508). This method can effectively identify corn seedlings in complex backgrounds, providing technical support for accurate detection and counting of corn seedlings in multiple weather conditions.

Abstract in Chinese

准确统计玉米苗数是评估玉米生长状况的关键，本研究为应对复杂田间环境玉米幼苗检测计数难的问题，提出改进MEI-YOLOv11模型。通过引入MANet、EUCB模块及Inner-SIoU损失函数，增强复杂环境下特征提取与小目标识别能力。结果表明，该模型的mAP0.5、P和R分别达到97.0%、94.2%和95.7%，分别比YOLOv11提高2.8、2.7和2.4个百分点。参数量和推理时间分别只增加了1.28M和0.4ms，检测精度优于其他检测模型。多天气计数准确率均在90%以上，晴天准确率最高，为91.23% （RMSE=4.5044, R²=0.8508）。该方法能够有效识别复杂背景下玉米幼苗，为多天气玉米幼苗准确检测计数提供技术支持。