近红外光谱的不同产地柑橘无损鉴别方法

柑橘是世界第一大水果。不同产地的柑橘内部品质和价格有所不同，但其外观差别较小，外行人较难通过肉眼实现准确鉴别分析。DNA标记法与仪器分析操作复杂、成本较高，且对样品具有破坏性，无法实现快速无损分析，影响了产品的二次销售。近红外光谱技术是一种快速无损的新型检测手段，可以用于不同产地农产品的鉴别分析。由于柑橘皮对光谱的干扰较大，导致现阶段柑橘产地无损鉴别研究匮乏。此外柑橘体积较大，因此需要对光谱采样点进行优化。为此，基于近红外光谱技术与化学计量学方法，提出了一种用于不同产地柑橘无损鉴别的新方法。使用近红外光谱仪得到了120个来自云南、湖南、广西武鸣、广西来宾的沃柑漫反射光谱数据。采用单一和组合光谱预处理方式以消除光谱中的多种干扰；采用主成分分析方法对数据进行降维处理，并以此作为输入值结合Fisher线性判别分析方法构建柑橘产地鉴别模型，并与主成分分析模型进行对比。此外，考察了不同光谱采样位置（赤道线4个采集点、果梗部以及果顶部）对结果的影响。结果表明：主成分分析方法结合优化光谱预处理的方法不能实现不同产地柑橘的准确鉴别分析，最优鉴别率仅为5%；而采用主成分分析-Fisher线性判别分析方法，利用赤道线4个点的平均光谱结合去偏置校正或多元散射校正预处理方法可实现不同产地柑橘的100%鉴别分析；采用主成分分析-Fisher线性判别分析对6个点的平均光谱数据进行处理时，采用原始光谱便可实现不同产地柑橘的100%鉴别分析。为此，通过对光谱预处理方法以及光谱采集点的优化，利用主成分分析-Fisher线性判别分析方法即可建立准确的柑橘产地鉴别模型，为不同产地柑橘的快速鉴别提供了新途径，为后续各种柑橘类水果的鉴别分析提供了参考。

A Nondestructive Identification Method of Producing Regions of Citrus Based on Near Infrared Spectroscopy

Citrus is one of the popular fruits in the world. There are differences in internal quality and price of different-regions citrus. However, the appearance differences are small, and it is difficult for laypeople to identify by appearance. Methods such as DNA labeling and instrumental analysis are complex in operation and destructive to samples, which cannot achieve rapid and non-destructive analysis, affecting the secondary sales of products. Near-infrared spectroscopy is a fast and nondestructive detection method that can be used to identify different-regions agricultural products. Due to the large interference of citrus peel on the spectra, there is a lack of nondestructive identification of citrus origin. Besides, citrus is large. Therefore, it is necessary to optimize the spectral collection points. This paper proposed a new method for nondestructive identification of different-regions citrus based on near infrared spectroscopy and chemometrics. The diffuse reflectance spectra of 120 fertile oranges from Yunnan, Hunan, Wuming and Laibin of Guangxi were obtained by near-infrared spectroscopy. Single and combined spectral pretreatment was used to eliminate multiple interferences in the spectra. The principal component analysis method was used to reduce the data dimension, which was used as the input value. Combining with Fisher linear discriminant analysis method, the citrus origin identification model was obtained and compared with the principal component analysis model. In addition, the effects of different spectral collection locations (4 collection points along the equator, top and bottom) on the results were investigated. The results showed that the principal component analysis method combined with the optimized spectral pretreatment method could not accurately identify different-regions citrus, and the best identification accuracy was 5%. When the principal component analysis-Fisher linear discriminant analysis was used, the average spectra of 4 collection points along the equator combined with De-bias correction or multivariate scattering correction pretreatment method could achieve 100% identification analysis of different-regions citrus. Furthermore, the average spectra of 6 collection points combined with raw data could achieve 100% identification analysis of different-regions citrus. Therefore, by optimizing the spectral pretreatment methods and spectral collection points, the accurate identification model of different-regions citrus can be established by using principal component analysis-Fisher linear discriminant analysis method, which provides a new method for rapid identification of different-regions citrus.