摘要
针对自然环境下番茄叶片存在的复杂背景和密集遮挡情况,提出一种改进的YOLOv5模型,用于实时检测自然环境下番茄叶片的病害。首先,使用RepVGG模块代替YOLOv5中主干网络的卷积层,改善主干网络的特征提取能力,减少模型的内存占用,加速模型的推理速度;其次,在颈部的C3模块中引入注意力机制模块CBAM,提高模型在复杂背景下对番茄患病叶片的检测精度以及对遮挡目标的识别率;最后,引用新的损失函数SIoU,加速模型的收敛速度并降低模型的损失值。研究结果表明,相比于原YOLOv5模型,改进模型的平均精度提升3.0个百分点,平均精度高达98.9%,说明改进模型在自然环境下对番茄患病叶片的检测更具优势。
Aiming at the complex background and dense occlusion of tomato leaves in natural environment,an improved YOLOv5 model was proposed for real-time detection of tomato leaf diseases in natural environment.Firstly,the RepVGG module was used to replace the convolutional layer of the backbone network in YOLOv5,which improved the feature extraction capability of the backbone network,reduced the memory occupation of the model and accelerated the reasoning speed of the model.Secondly,the attention mechanism module CBAM was introduced into C3 module in neck part to improve the detection accuracy of tomato diseased leaves and the recognition rate of shielded targets in the complex background.Finally,a new loss function SIoU was introduced to accelerate the convergence speed of the model and reduce the loss value of the model.The research results showed that compared with the original YOLOv5 model,the average precision of the improved model increased by three percentage points,and the average accuracy was as high as 98.9%,indicating that the improved model was more advantageous in the detection of tomato diseased leaves in natural environment.
作者
李仁杰
宋涛
高婕
李东
高鹏
李炳鑫
杨坡
LI Renjie;SONG Tao;GAO Jie;LI Dong;GAO Peng;LI Bingxin;YANG Po(School of Electronic and Information Engineering,Hebei University of Technology,Tianjin 300000,China)
出处
《江苏农业学报》
CSCD
北大核心
2024年第6期1028-1037,共10页
Jiangsu Journal of Agricultural Sciences
基金
河北省重点研发计划项目(22370701D)。
