基于BP神经网络的大豆叶片病害诊断模型的研究

本文以大豆叶片为研究对象,主要针对大豆灰斑病、霜霉病和细菌性斑点病进行诊断。首先,在东北农业大学教育部大豆生物学重点实验室的实验基地培育灰斑病、霜霉病和细菌性斑点病的纯正样本,然后通过对病斑特征的分析,确定病斑特征与病害种类的关系,建立大豆叶片病害的BP神经网络诊断模型。测试结果表明,针对轻度病害,灰斑病、霜霉病、细菌性斑点病和其它病害的识别精度分别为88.75%,87.50%,87.50%,85.00%;中度病害识别精度分别为91.25%,90.00%,91.25%,88.75%;重度病害识别精度分别为93.75%,92.50%,93.75%,92.50%。

Research on Soybean Leaf Disease Diagnosis Based on BP Network

Aiming at gray leaf spot, downy mildew and bacterial spot disease for soybean leaves, a diagnosis method is proposed in this paper. First, the pure samples of gray leaf spot, downy mildow and bacterial spot disease for soybean leaves are forstered in the northeast agricultural university experimental base of soybean biology key laboratory of the ministry of education. Then, analyze the characteristics of disease spot to determine the relationships of disease spot characteristics and damage types to establish soybean leaf diseases diagnosis model based on BP neural network. Test results show that the recognition accuracy for mild disease （gray leaf spot, downy mildew, bacterial spot disease and other diseases） are 88.75%,87.50%,87.50%,85.00%, the recognition accuracy for moderate disease are 91.25%,90.00%,91.25%,88.75%, and the recognition accuracy for severe disease are 93.75%,92.50%,93.75%,92.50%.