摘要
叶片枯萎是玉米生长过程中最常见的症状之一,利用视觉传感与模式识别技术实现玉米叶枯萎的自动检测可极大提高玉米的产量和生产效率,是智慧农业发展的重要方向。针对玉米叶遮挡场景下的枯萎检测问题,通过YOLOv5单阶段目标检测框架对叶片进行建模,并结合通道和空间注意力机制对模型架构进行改进,增强其骨干网络的特征提取能力,提高小目标玉米叶枯萎的检测精度。试验结果表明,提出的方法在Stewart_NLBimages_2019数据集上检测性能优于YOLOv5,检测平均精度均值达到86.84%,具有广阔的应用前景。
Leaf blight is one of the most common symptoms during the growth of corn.Utilizing computer vision and pattern recognition techniques to achieve automatic detection of corn leaf blight can significantly improve the corn yield and production efficiency,making it an important direction in smart agriculture.To address the challenge of blight detection in the presence of occluded corn leaves,we employ the YOLOv5 single-stage object detection framework to model leaves.Furthermore,we enhance the feature extraction capabilities in the backbone network through the integration of channel and spatial attention mechanisms.This improvement enhances the detection accuracy of small target areas affected by corn leaf blight.Experimental results demonstrate that the proposed method outperforms YOLOv5 in terms of detection performance on the Stewart_NLBimages_2019 dataset,achieving an average precision of 86.84%,which shows its great prospect in widespread applications.
作者
方影
王亮亮
FANG Ying;WANG Liangliang(Mengcheng County Agricultural Mechanization Technology Promotion Service Station,Mengcheng 233500,China;School of Computer Science and Artificial Intelligence,Wuhan University of Technology,Wuhan 430070,China;Chongqing Research Institute,Wuhan University of Technology,Chongqing 401120,China)
出处
《农机使用与维修》
2024年第3期5-8,共4页
Agricultural Machinery Using & Maintenance
基金
重庆市自然科学基金(CSTB2022NSCQ-M)。
作者简介
方影(1972—),女,安徽蒙城人,大专,工程师,研究方向为农业工程;通讯作者:王亮亮(1987—),男,山东临沂人,博士,讲师,研究方向为机器视觉。
