摘要
基于高光谱成像及人工神经网络技术对玉米含水率进行了检测。检测波长为450～900nm,由玉米粒反射光谱图像获取反映其含水率的光谱特征波长。利用人工神经网络建立了玉米粒含水率的预测模型,模型相关系数达到0.98。对含水率预测结果的误差最大绝对值为2.1182,最小绝对值为0.0024。相对误差绝对值的平均值为0.3090,结果表明利用高光谱图像技术对玉米含水率进行无损检测是可行的。
Water content is an important quality attributer.It was investigated that water content in corn was detected based on hyperspectral imaging and neural network.The detection wavelengths ragion between 450 and 900 nm.The spectrum features wavelengths for predicting the water content in corn were obtained by scatting spectral images.Subsequently,artificial neural network was used for developing a prediciton model to predict water content in corn.The prediction results showed that the maximal absolute value of error was 2.1182,the minimal absolute value of error was 0.0024,the average was 0.3090.Therefore,the hyperspectral imaging is an effective method for nondestructive assessing the water content in corn.
出处
《包装与食品机械》
CAS
2010年第6期1-4,共4页
Packaging and Food Machinery
基金
国家科技支撑计划(2008BADA8B04)
关键词
玉米
高光谱图像
含水率
神经网络
corn
hyperspectral imaging
water content
neural network
作者简介
李江波（1982-），男，博士生，研究方向为基于计算机视觉技术的水果表面缺陷检测。
通讯作者：饶秀勤（1968-），男，博士，副教授，研究方向为农产品无损检测。通讯地址：310029杭州浙江大学生物系统工程与食品科学学院，E-mail：xqrao@zju．edu．cn。
