Rapid Determination Of Fruit Hardness And SSC Of Xiangshui Pear By Hyperspectral During Storage

As one of the important industries in Haiyuan County,Ningxia,Xiangshui pear has an important impact on the local economic development.Hardness is the most intuitive indicator reflecting the maturity of pears and one of the important criteria for maturity judgment.The soluble solid content(SSC)has an important influence on the taste and flavor of the fruit during the ripening process of Xiangshui pear,and it is an important parameter to reflect the quality of pear.Based on Visible near infrared reflection(VisNIR)and near infrared reflectance spectroscopy(NIR)was used to analyze the hardness and soluble solid content of Xiangshui Pear,combined with Savitzky-Golay(SG)and other preprocessing methods,competitive adaptive weighting algorithm(CARS)and other characteristic wavelength extraction methods,respectively established the partial least squares(PLSR)model of the hardness and soluble solids content of Xiangshui pears.At the same time,principal component analysis discriminant analysis(PCADA)and other three discriminant models were used to distinguish the fragrant pear with different storage periods.The specific research contents are as follows:(1)The hyperspectral prediction model of Xiangshui pear fruit hardness was established.Five preprocessing methods and five characteristic wavelength extraction algorithms were used to process the spectral data to establish Vis-NIR and VIR hyperspectral PLSR predictions of Xiangshui pear hardness.model.The results show that in the range of Vis-NIR band,Normalize method is the optimal preprocessing method.and 10 characteristic wavelengths are extracted by VCPA algorithm and the PLSR model established after extracting the characteristic wavelengths based on the VCPA algorithm PLSR model based on CARS algorithm has the best effect,which is 16.62%higher than the prediction effect of the original spectral model;In the range of NIR bands,SNV method is the optimal preprocessing method,and 23 characteristic wavelengths are extracted by CARS algorithm.PLSR model based on CARS algorithm has the best effect,which is 41.86%higher than the prediction effect of the original spectral model.(2)The hyperspectral prediction model of the soluble solids of Xiangshui pear fruit was established.Five preprocessing methods and five characteristic wavelength extraction algorithms were used to process the spectral data to establish Vis-NIR and VIR hyperspectral PLSR predictions of Xiangshui pear hardness.model.The results show that in the range of Vis-NIR band,MSC method is the optimal preprocessing method.and 61 characteristic wavelengths are extracted by iVISSA algorithm and the PLSR model established after extracting the characteristic wavelengths based on the iVISSA algorithm PLSR model based on CARS algorithm has the best effect,which is 23.93%higher than the prediction effect of the original spectral model;In the range of NIR bands,MSC method is the optimal preprocessing method,and 11 characteristic wavelengths are extracted by VCPA algorithm.PLSR model based on VCPA algorithm has the best effect,which is 20.62%higher than the prediction effect of the original spectral model.(3)Study on the storage period discriminant model of perfume pear based on hyperspectral technology,using PCA-DA,PLS-DA,SIMCA three models to discriminate and analyze the storage period of Xiangshui pear.The results show that the PCA-DA and PLS-DA models have a better discrimination effect and a relatively high accuracy rate;in the Vis-NIR band,the CARS method is used to extract 23 characteristic wavelengths,and the CARS algorithm is used to extract the characteristic wavelengths.The established PCA-DA model prediction set discriminating accuracy rate is 96%;in the NIR band,121 characteristic wavelengths are extracted using the iVISSA method,and the PLSDA model established based on the iVISSA algorithm processing and extracting the characteristic wavelengths has a discriminant accuracy rate of 98%.