摘要
自动和准确地估计病害的严重度对病害管理和产量损失预测至关重要。针对传统病害严重度估算步骤复杂且低效,难以实现在田间场景下精准估算问题,提出了一种基于混合扩张卷积和注意力机制改进UNet(Mixed dilated convolution and attention mechanism optimized UNet,MA-UNet)的病害严重度估算方法。首先,针对病斑尺寸不一、形状不规则问题,提出混合扩张卷积块(Mixed dilation convolution block,MDCB)增加感受野并保持病斑信息的连续性,提升病斑分割精度。其次,为了克服复杂背景的影响,利用注意力机制(Attention mechanism)对空间维度和通道维度进行相关性建模,获得每个像素类内响应和通道间的依赖关系,缓解背景对网络学习带来的影响。最后,计算病害分割图中病斑像素与叶片像素的比率来获得严重度。基于田间条件下收集的黄瓜霜霉病和白粉病图像进行了验证,并与全卷积网络(Fully convolutional network,FCN)、SegNet、UNet、PSPNet、FPN、DeepLabV3+进行比较。结果表明,MA-UNet优于比较方法,能够满足复杂环境下健康叶片和病斑的分割需求,平均交并比为84.97%,频权交并比为93.95%。基于MA-UNet分割结果估计黄瓜叶部病害严重度的决定系数为0.9654,均方根误差为1.0837%。该研究可为人工智能在农业中快速估计和控制病害严重度提供参考。
Automatic and accurate estimation of disease severity is critical for disease management and yield loss prediction.Traditional disease severity estimation steps are complicated and inefficient,which makes it challenging to achieve accurate estimation in practical scenarios.A disease severity estimation method was proposed based on mixed dilated convolution and attention mechanism to improve UNet(MA-UNet).Firstly,to solve the problem of different sizes and irregular shapes of lesions,the mixed dilation convolution block(MDCB) was proposed to increase the receptive field and maintain the continuity of lesion information to improve the accuracy of lesion segmentation.Secondly,to overcome the influence of complex background,the attention mechanism(AM) was used to model the correlation between the spatial and channel dimensions.It can obtain the response within each pixel class and the dependency between channels to alleviate the background's influence on network learning.Finally,the ratio of diseased lesion pixels to leaf pixels in the disease segmentation map was calculated to obtain the severity.It was validated based on cucumber downy mildew and powdery mildew images collected under field conditions and compared with fully convolutional network(FCN),SegNet,UNet,PSPNet,FPN,and DeepLabV3+.MA-UNet can meet the segmentation requirements of leaves and lesions in complex environments,with a mean intersection over union 84.97% and a value of frequency-weighted intersection over union value of 93.95%.Moreover,it can accurately estimate the severity of cucumber leaf diseases,the correlation coefficient was 0.965 4,and the RMSE was 1.083 7%.The results showed that MA-UNet outperformed the comparison methods in refining lesion segmentation and accurately estimating disease severity.The research result can provide a reference for artificial intelligence to estimate and control disease severity in agriculture rapidly.
作者
李凯雨
朱昕怡
马浚诚
张领先
LI Kaiyu;ZHU Xinyi;MA Juncheng;ZHANG Lingxian(College of Information and Electrical Engineering,China Agricultural University,Beijing 100083,China;Institute of Environment and Sustainable Development in Agriculture,Chinese Academy of Agricultural Sciences,Beijing 100081,China)
出处
《农业机械学报》
EI
CAS
CSCD
北大核心
2023年第2期231-239,共9页
Transactions of the Chinese Society for Agricultural Machinery
基金
国家自然科学基金项目(62176261)。
关键词
黄瓜病害
病害严重度
扩张卷积
注意力机制
语义分割
cucumber disease
disease severity
dilation convolution
attention mechanism
semantic segmentation
