摘要
苹果的轻微损伤部位容易被病原微生物入侵,导致自身和周围果实腐烂。早期轻微损伤的检测能有效降低损失。本研究采用单积分球系统结合反向倍增法测量1450~1800 nm波段下极品富士苹果的光学参数,采用多元散射校正和标准正态变量对数据进行预处理。基于1450~1800 nm的吸收系数和约化散射系数,结合主成分分析和概率神经网络(PNN)建立了苹果轻微损伤的判别模型。构建的模型对无损伤样品判断准确率达到96%以上,当PNN密度分度值(Spread)小于0.7时,吸收系数对有损伤样品的判断准确率为100%。实验结果说明了苹果在1450~1800 nm范围内的光学性质能够用于判断苹果早期轻微损伤,为光学参数检测水果损伤提供应用前景。
The positions of tiny bruising apples are easily invaded by pathogenic microorganisms,causing themselves and surrounding fruits to rot.The early detection can effectively reduce the loss from tiny bruise.In this study,the single integrating sphere setup and the inverse adding-doubling（ IAD） method were used to investigate the optical properties of Fuji apples at1450~1800 nm.The data of optical properties were preprocessed by the multiplicative scatter correction and standard normal variables methods.A discriminate model that combines principal component analysis with Probabilistic neural networks（ PNN） to judge and predict the bruising of apples was established based on the absorption coefficient and the reduced scattering coefficient of 1450~1800 nm.The accuracy of the model for no bruising judgment was above 96%.The accuracy of the model of absorption coefficients for early bruising judgment was 100% when spread was less 0.7.The experimental results showed optical properties can judge the early tiny bruising of apples and provide application prospect of detecting the bruise of fruits.
出处
《食品工业科技》
CAS
CSCD
北大核心
2017年第18期258-263,共6页
Science and Technology of Food Industry
基金
农林高校融合型物理实验课程体系的改革与实践
湖北省教学研究项目(2012172)
关键词
早期轻微损伤
苹果
单积分球
光学性质
PNN网络
early tiny bruise
apples
integrating spheres
optical properties
probabilistic neural networks
