[目的/意义]炭疽病(anthracnose)作为油茶生长过程中重要的病害,其严重程度的精准判定对于精准施药和科学管理具有重大意义。本研究提出了一种改进YOLACT(You Only Look At CoefficienTs)分级模型Camellia-YOLACT,旨在实现对油茶叶片炭...[目的/意义]炭疽病(anthracnose)作为油茶生长过程中重要的病害,其严重程度的精准判定对于精准施药和科学管理具有重大意义。本研究提出了一种改进YOLACT(You Only Look At CoefficienTs)分级模型Camellia-YOLACT,旨在实现对油茶叶片炭疽病感染严重程度的自动、高效判定。[方法]首先在YOLACT主干网络部分使用Swin-Transformer来进行特征提取。Transformer架构的自注意力机制拥有全局感受野及移位窗口等特性,有效地增强了模型的特征提取能力;引入加权双向特征金字塔网络,融合不同尺度的特征信息,加强模型对不同尺度目标的检测能力,提高模型的检测精度;在激活函数的选择上,采用非线性能力更强的HardSwish激活函数替换原模型的ReLu激活函数。由于HardSwish在负值区域不是完全截断,对于输入数据中的噪声具有更高的鲁棒性,自然环境下的图像有着复杂的背景和前景信息,HardSwish的鲁棒性有助于模型更好地处理这些情况,进一步提升精度。[结果和讨论]采用迁移学习方式在油茶炭疽病感染严重程度分级数据集上进行实验验证。消融实验结果表明,本研究提出的Camellia-YOLACT模型的mAP75为86.8%,较改进前提升5.7%;mAPall为78.3%,较改进前提升2.5%;mAR为91.6%,较改进前提升7.9%。对比实验结果表明,Camellia-YOLACT在精度和速度方面表现均好于SOLO(Segmenting Objects by Locations),与Mask R-CNN算法相比,其检测速度提升了2倍。在室外的36组分级实验中进一步验证了Camellia-YOLACT模型的性能,其对油茶炭疽病严重程度的分级正确率达到了94.4%,K值平均绝对误差为1.09%。[结论]本研究提出的Camellia-YOLACT模型在油茶叶片和炭疽病病斑分割上具有较高的精度,能够实现对油茶炭疽病严重程度的自动分级,为油茶病害的精准防治提供技术支持,进一步推动油茶炭疽病诊断的自动化和智能化。展开更多
The threat posed to crop production by pests and diseases is one of the key factors that could reduce global food security.Early detection is of critical importance to make accurate predictions,optimize control strate...The threat posed to crop production by pests and diseases is one of the key factors that could reduce global food security.Early detection is of critical importance to make accurate predictions,optimize control strategies and prevent crop losses.Recent technological advancements highlight the opportunity to revolutionize monitoring of pests and diseases.Biosensing methodologies offer potential solutions for real-time and automated monitoring,which allow advancements in early and accurate detection and thus support sustainable crop protection.Herein,advanced biosensing technologies for pests and diseases monitoring,including image-based technologies,electronic noses,and wearable sensing methods are presented.Besides,challenges and future perspectives for widespread adoption of these technologies are discussed.Moreover,we believe it is necessary to integrate technologies through interdisciplinary cooperation for further exploration,which may provide unlimited possibilities for innovations and applications of agriculture monitoring.展开更多
文摘[目的/意义]炭疽病(anthracnose)作为油茶生长过程中重要的病害,其严重程度的精准判定对于精准施药和科学管理具有重大意义。本研究提出了一种改进YOLACT(You Only Look At CoefficienTs)分级模型Camellia-YOLACT,旨在实现对油茶叶片炭疽病感染严重程度的自动、高效判定。[方法]首先在YOLACT主干网络部分使用Swin-Transformer来进行特征提取。Transformer架构的自注意力机制拥有全局感受野及移位窗口等特性,有效地增强了模型的特征提取能力;引入加权双向特征金字塔网络,融合不同尺度的特征信息,加强模型对不同尺度目标的检测能力,提高模型的检测精度;在激活函数的选择上,采用非线性能力更强的HardSwish激活函数替换原模型的ReLu激活函数。由于HardSwish在负值区域不是完全截断,对于输入数据中的噪声具有更高的鲁棒性,自然环境下的图像有着复杂的背景和前景信息,HardSwish的鲁棒性有助于模型更好地处理这些情况,进一步提升精度。[结果和讨论]采用迁移学习方式在油茶炭疽病感染严重程度分级数据集上进行实验验证。消融实验结果表明,本研究提出的Camellia-YOLACT模型的mAP75为86.8%,较改进前提升5.7%;mAPall为78.3%,较改进前提升2.5%;mAR为91.6%,较改进前提升7.9%。对比实验结果表明,Camellia-YOLACT在精度和速度方面表现均好于SOLO(Segmenting Objects by Locations),与Mask R-CNN算法相比,其检测速度提升了2倍。在室外的36组分级实验中进一步验证了Camellia-YOLACT模型的性能,其对油茶炭疽病严重程度的分级正确率达到了94.4%,K值平均绝对误差为1.09%。[结论]本研究提出的Camellia-YOLACT模型在油茶叶片和炭疽病病斑分割上具有较高的精度,能够实现对油茶炭疽病严重程度的自动分级,为油茶病害的精准防治提供技术支持,进一步推动油茶炭疽病诊断的自动化和智能化。
基金supported by National Key Research and Development Program of China(Grant No.2022YFC2602100)Chinese Academy of Inspection and Quarantine(2022JK38).
文摘The threat posed to crop production by pests and diseases is one of the key factors that could reduce global food security.Early detection is of critical importance to make accurate predictions,optimize control strategies and prevent crop losses.Recent technological advancements highlight the opportunity to revolutionize monitoring of pests and diseases.Biosensing methodologies offer potential solutions for real-time and automated monitoring,which allow advancements in early and accurate detection and thus support sustainable crop protection.Herein,advanced biosensing technologies for pests and diseases monitoring,including image-based technologies,electronic noses,and wearable sensing methods are presented.Besides,challenges and future perspectives for widespread adoption of these technologies are discussed.Moreover,we believe it is necessary to integrate technologies through interdisciplinary cooperation for further exploration,which may provide unlimited possibilities for innovations and applications of agriculture monitoring.