基于BP神经网络小麦含水量的近红外检测方法

NEAR INFRARED DETECTION OF WHEAT WATER CONTENT BASED ON BP NEURAL NETWORK

为了实现小麦籽粒水分含量的快速、准确测定,以小麦含水量的国标测定值为BP神经网络的目标向量,采用近红外分析仪扫描小麦籽粒光谱图并进行预处理,使用主成分分析对预处理后的光谱数据降维,将其作为BP神经网络的输入向量,采用BP神经网络对小麦的含水量进行预测.结果表明:国标法测定值与BP神经网络结合近红外法测定值之间的T检验结果为P=0.52>0.05,两种方法测定结果无显著性差异.采用BP神经网络对小麦水分的预测值与国标测定值之间的R^2为0.999,可以用来预测小麦的水分含量.

In order to achieve rapid and accurate determination of water content in wheat grains, we pretreated the spectra of wheat grains scanned by near infrared spectrometer, processed the spectral data by principal component analysis to reduce dimension, and predicted the wheat water content by using the BP neutral network which selected the national standard measurement value of wheat water content as the target vector and took the processed spectral data as the input vector. The results showed that the T detection result between the national standard measurement value and the determination value obtained by the BP neutral network and near infrared spectroscopy was P=-0.52, which indicated that the detection results of two methods had no significant difference; and R2 between the prediction value of the BP neutral network and the national standard measurement value was 0.999, which indicated that the BP neutral network could be used for predicting the wheat water content.