一种茶叶病害的深度学习检测算法

A Deep Learning Detection Algorithm for Tea Diseases

提出了一种改进的基于深度学习的茶叶病害目标检测算法。该方法在网络模型中添加坐标注意力机制,使模型细化特征,更加关注茶叶病害信息,从而抑制树枝、杂草等一些背景因素的干扰;选用CIoU作为模型的损失函数以提高定位能力;同时,通过聚类的方法对数据集中的目标边框进行重新优化,以获得更精准的先验框;并建立包含6种病害的茶叶病害数据集,解决了病害图像数据匮乏的问题。与其他算法对比实验结果表明,所提出的算法在多个指标上均有较好的表现,可为茶叶病害智能化诊断提供高效的解决方案。

An improved deep learning algorithm based tea disease target detection was proposed.A coordinate attention mechanism was incorporated into the network,which could enable the model to refine features and focus more on disease information,thereby suppress the interference from background factors such as branches and weeds.CIoU was selected as the loss function of the model to improve the localization capabilities.Simultaneously,the target bounding boxes in the dataset was optimized through clustering techniques to obtain more accurate prior boxes.To address the issue of insufficient disease image data,a tea diseases dataset comprising six disease types was established.Experimental results showed that,compared to other algorithms,the presented method could exhibit superior performance across multiple metrics,and provide an efficient solution for the intelligent diagnosis of tea diseases.