基于人工神经网络的苹果可溶性固形物无损检测  

Nondestructive Examination of Apple Soluble Solids Content Based on Artificial Neural Network

作　　者：代芬[1,2] 洪添胜[1,2] 尹令[1,2] 代秋芳[1,2] 张昆[1,2] 

机构地区：[1]华南农业大学南方农业机械与装备关键技术省部共建教育部重点实验室,广州510642 [2]华南农业大学工程学院,广州510642

出　　处：《农机化研究》2011年第10期134-137,共4页Journal of Agricultural Mechanization Research

基　　金：国家自然科学基金项目(30871450);华南农业大学校长基金项目(4500-k09173)

摘　　要：采集了60个苹果在400～1 100nm范围内的可见-近红外漫反射光谱,然后使用连续投影算法将光谱变量进行压缩,最后采用BP神经网络建立了苹果糖度的预测模型。实验表明,连续投影算法从400～1 100nm范围提取出25个优选波长参与建模,有效简化了模型结构。BP神经网络模型对苹果糖度的预测相关系数达到0.853,预测均方根误差为1.303 0。结果表明,基于近红外光谱的苹果糖度无损检测是可行的。The reflectance spectra of 60 APPLE SAMPLES were collected.Then the spetra were composed by Successive Projections Algorithm.finally,the Artificial Neural Network model of apple sugar content was built.As a result,25 spectra variables were derived from the 400-1100nm spectra by Successive Projections Algorithm and the ANN model based on these 25 variables produced RP=0.853 and RMSEP=1.3030.it was demonstrated that the Nondestructive Examination of Apple Sugar based on near infrared spectrum is feasible.

关 键 词：苹果 近红外光谱 无损检测 人工神经网络 连续投影算法 糖度 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] S126[自动化与计算机技术—控制科学与工程]