基于BP神经网络的大米含水量近红外检测方法  

A Near-Infrared Detection Method of Moisture Content in Rice Based on BP Neural Network

作　　者：孙永海[1] 万鹏[1] 于春生[1] 

机构地区：[1]吉林大学生物与农业工程学院,长春130022

出　　处：《中国粮油学报》2008年第6期193-197,共5页Journal of the Chinese Cereals and Oils Association

基　　金：吉林省人才开发基金(200605);吉林大学"种子基金"(419070402418)

摘　　要：以大米含水量的国标检测方法为基础,采用近红外检测仪对大米的含水量进行检测,获取近红外光在各个波长处的反射光检测值,并对反射光检测值与大米含水量之间的相关性进行分析。将相关性较高的近红外反射光检测值作为输入值,采用BP神经网络对大米的含水量进行检测,并与近红外检测仪的回归方程的检测结果进行对比。试验结果表明,采用近红外反射光检测值输入BP神经网络检测大米含水量的方法准确率可达到96.67%;而采用近红外检测仪的回归方程进行检测的准确率为90.00%。Based on the national standard detection method of rice moisture content,rice moisture contents were detected by a near infrared spectroscopy.The detection data were gained in every wavelength of the near infrared spectroscopy,and the correlation between the detection data and the rice moistures was analyzed.The highly correlative data of the wavelengths were regarded as input data for BP Neural Networks and moisture contents were detected using the BP Neural Networks,which was compared with the result detected using the regression equation of the near infrared system.Results show that the accuracy of detecting rice moisture content based on the BP Neural Networks is about 96.67%,and the accuracy of detecting rice moisture by using the regression equation of the near infrared system is only 90.00%.

关 键 词：食品科学技术基础学科 近红外检测 BP神经网络 含水量 大米 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] U472.9[自动化与计算机技术—控制科学与工程]