多光谱成像技术诊断植物病虫害的人工神经网络模型

Multispectral imaging techniques diagnosing plant diseases and insect pests using artificial neural networks

为了实现可靠的植物病虫害诊断,提出把人工神经网络和多光谱成像技术结合的方法,并将该方法用于常见的三种黄瓜病害的识别研究。在此基础上,实验采用窄带多光谱成像技术获取患病黄瓜叶面的14个可见光通道和近红外通道、全色通道的多光谱图像。利用BP网络对病斑样本的光谱信息进行学习分类。和14通道训练结果比较,增加850nm的近红外通道和全色通道,使网络的训练时间缩短、预测能力提高。实验结果表明,这种方法对植物进行快速、准确和非破坏性诊断提供可靠的技术支持。

For a reliable diagnosis of plant diseases and insect pests, artificial neural networks and mutispectral imaging technique are proposed to diagnosise three cucumbers diseases. In the experiment, the cucumbers multispectral images of 14 visible lights channels, near infrared charmel and panchromatic channel are captured using narrow-band multispectral imaging system, and classified by the BP neural network. The coefficients of output and prediction are obtained. The result shows that the method realizes good accuracy in the cucumber diseases diagnosis.