期刊文献+

基于改进YOLOv8的水稻叶片病害检测方法研究

Study on Detection Method of Rice Leaf Diseases Based on Improved YOLOv8
下载PDF
导出
摘要 针对当前农业领域中水稻叶片病害检测过程中存在的人工识别精度不高和效率低下的问题,提出一种基于改进的YOLOv8模型的水稻叶片病害检测方法。首先分析了稻瘟病、白叶枯病和褐斑病三种病害图像构成的数据集的相关性,在原有的YOLOv8结构中融合OREPA技术,获得轻量化和更加丰富的梯度流信息,进而改进模型性能以及降低计算资源需求,然后在水稻病害数据集上将改进后的模型与其他模型进行训练对比。实验结果表明,改进的YOLOv8模型在处理多样化的水稻叶片病害图像时表现出更高的准确性,其检测精确度、召回率、m AP50以及mAP50-95性能指标分别为97.5%、91.4%、96%和76.3%。该方法不仅为水稻叶片病害的自动化检测提供了技术方案,也为其他农作物病害检测的研究提供了参考,对促进可持续农业发展具有重要意义。 In view of the problems of low accuracy and efficiency in manual identification during rice leaf disease detection in the current agricultural field,a detection method for rice leaf diseases based on an improved YOLOv8 model was proposed.Firstly,the correlation of datasets composed of images of rice blast,bacterial blight,and brown spot was analyzed.OREPA technology was integrated into the original YOLOv8 structure to obtain lightweight and richer gradient flow information,thus improving model performance and reducing computing resource requirements.Then the improved model was trained and compared with other models on the rice disease dataset.The test results showed that the improved YOLOv8 model exhibited higher accuracy in processing diversified rice leaf disease images,and its detection precision,recall rate,mAP50 and mAP50-95 performance indexes were 97.5%,91.4%,96%,and 76.3%respectively.This method not only provides a technical scheme for the automatic detection of rice leaf diseases,but also offers references for other crop disease detection research,which is of great significance in promoting sustainable agricultural development.
作者 陆军 谢锋 LU Jun;XIE Feng(Publicity Office of the CPC Nanning Vocational and Technical University Committee,Nanning,Guangxi 530008,China;School of Artificial Intelligence,Nanning Vocational and Technical University,Nanning,Guangxi 530008,China)
出处 《广西农学报》 2024年第4期27-35,42,共10页 Journal of Guangxi Agriculture
基金 广西创新驱动重大专项(2020AA24002AA)。
关键词 YOLOv8 水稻叶片病害 精准农业 OREPA YOLOv8 Rice leaf disease Precision agriculture OREPA
  • 相关文献

参考文献1

二级参考文献35

共引文献10

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部