基于拉曼光谱的稻叶瘟病害程度划分研究

随着植保变量喷施作业机械的研究和应用,急需一种高效的病害程度识别技术。为此,针对水稻稻叶病运用拉曼光谱仪采集正常及受病害叶片的光谱特性,通过绘制折线图及受试者工作特征曲线进行水稻受病害程度分析,并运用动量因子BP神经网络优化算法,建立了寒地水稻稻叶瘟的病害程度检测模型。结果表明:优化的BP神经网络算法网络预测集的均方误差为0. 002 409 6、相关系数为0. 998 2。该方法可以较好地区分水稻正常叶片、稻叶瘟重度和轻度叶片,是一种高效的病害程度识别技术。

Study on the Degree of Rice Leaf Blast Disease Based on the Raman Spectroscopy

With the research and application of plant protection variable spraying machine,an efficient identification technology of disease degree is urgently needed.In this paper,the spectral characteristics of normal and diseased leaves were collected by Raman spectroscopy for rice leaf disease.By drawing the line chart and subject operating characteristic curve to analyse the degree of rice disease,and use the optimization algorithm of momentum factor BP neural network to establish a model for detecting the disease degree of rice leaf blast in cold region.The results show that the mean square error and correlation coefficient of the optimized BP neural network prediction set is 0.0024096and 0.9982respectively.This method can be used to distinguish normal rice leaf,rice leaf blast and light leaf.It is an efficient identification technology of disease degree.