شناسایی آسیب یخزدگی میوه پرتقال رقم تامسون با استفاده از روشهای طیفسنجی تبدیل فوریه-فروسرخ و تصویربرداری فراطیفی. (Persian)

Introduction: Orange is a tropical and semi-tropical product sensitive to cold and freezing stress. Mostly, chilling occurs in the temperature range of 0oC and freezing occurs below 0oC. One of the most serious and significant damages to citrus fruits is freezing, which has a significant impact on the flavor and marketability of orange and orange juice. Changing the flavor of the product, even in juicing processes, can cause a sharp loss in the flavor of the produced juice. Fourier transform-infrared spectrometry (FT-IR) is known as a cost-effective method for detecting compounds in different materials. FT-IR shows the characteristics of molecular vibrations and functional groups in materials, so it can show the changes made in the structure of materials due to any temperature changes [10, 11]. In addition, the hyperspectral imaging method is also one of the non-destructive methods of examining surface defects with good accuracy, which has been the focus of various researchers today. Infrared hyperspectral imaging method was used to detect Escherichia coli in lettuce and the results showed that 4 different groups could be classified with more than 90% accuracy [12]. Therefore, in the present study, the aim was to investigate the ability of FT-IR and hyperspectral imaging to distinguish frozen oranges from unfrozen oranges. Despite the qualitative nature of the FT-IR method, the data of this method were also modeled quantitatively. Materials and methods: In this study, the frozen and unfrozen orange was investigated using FT-IR and hyperspectral imaging. In the FT-IR method 20 frozen and unfrozen orange peels samples and in the hyperspectral imaging method 18 frozen and unfrozen orange samples was used. At the first, oranges were examined based on the days of freezing, but the spectroscopy results (FT-IR and Hyperspectral) showed that there is no difference between the days of freezing. Therefore, as presented below, the samples were classified based on frozen and unfrozen by the supervised pattern recognition method of linear discriminant analysis (LDA). For that, first, pre-processing of the spectra using some methods of smoothing and noise reduction (moving averaging (MA), Savitzky-Golay (SG) and median filter), and normalization (Multiplicative Scatter Correction (MSC), Standard Normal Variate (SNV)) was performed. Studying the effect of the above five methods individually and six combined methods MA+MSC, MA+SNV, SG+MSC, SG+SNV, MF+MSC, and MF+SNV on the accuracy of LDA method was investigated. In this study, 75% of unfrozen and frozen samples were selected as calibration data and 25% as evaluation data. Then, in order to implement the LDA method, three linear, quadratic and Mahalanobis functions were used and classification was done accordingly. All the above steps were done in The Unscrambler X 10.4 software. Results and discussion: The results showed that the absorption peaks observed in different areas in all FT-IR spectra are also present in unfrozen samples, which are related to cellulose and lignin [17]. The broadest visible peak is in the approximately 3400 cm-1 wave number, which indicates the stretching vibrations of O-H groups in the structure of orange peel. The peak in 1063 cm-1 wave number indicates the C-O-H or C-O-R bond (alcohols or esters), and the peaks in 2924 cm-1 and 1447 cm-1 respectively indicate C-H stretching vibrations and aliphatic chains (- CH2- and -CH3-)[19]. In addition, two peaks are observed in 1739 cm-1 and 1634 cm-1 wave number, which are respectively caused by the presence of carbonyl groups such as ester and aliphatic and/or unsaturated aromatic compounds in the structure of orange peel [19, 20]. The main compound in the structure of orange peel in the FT-IR spectrum is lignin [21]. The absorption peaks in 400 cm-1 to approximately 1500 cm-1 wave number have undergone a fundamental change and the intensity of these peaks has been greatly reduced. Because peel of citrus fruits have secretory bags or in other words, glands containing oily substances (mainly contain peel essential oil), which burst due to freezing and the oily substance spreads on the peel. Spilling these substances will burn the skin cells located in the spaces between these glands and cause them to wrinkle [23]. Freezing causes water to escape and drying of the essential oil bags in the peel and the formation of white hesperidin crystals inside the damaged fruits. Therefore, the results of the FT-IR spectrum of orange samples after 2 to 12 days of freezing show that this method is capable of detecting frost damage in orange fruit from the samples taken from its peel. The results of FT-IR spectroscopy also showed that there is a difference between frozen and unfrozen samples in the range of 400 cm-1 to 1500 cm-1 wave number. According to the results in the linear and Mahalanobis methods, only for the MF+SNV combined preprocessing method, the correctly classified samples are 100 %. However, in the quadratic algorithm, it is 100% for all methods except MA, SG and MF. Considering that linear function is basically a simpler function than other functions, therefore LDA based on MF+SNV preprocessing and linear method was used as the basis of classification. The correctly classified samples (accuracy) in the evaluation category is around 92%, indicating the high ability of the model obtained from FT-IR spectroscopic data processed by the MF+SNV method to predict frozen and unfrozen of oranges. Conclusions: According to the results obtained in this study, after freezing, the peaks in the FT-IR spectrum, in the 400 cm-1 to a 1500 cm-1 wave number, underwent a fundamental change and the intensity of these peaks decreased drastically. Moreover, by applying the preprocessing methods by median filter and standard normal distribution (MF+SNV) of FT-IR spectroscopic data, it is possible to determine the frozen and unfrozen orange with high accuracy (100% training and 92% evaluation) recognized. The results of the hyperspectral imaging method showed that by applying the smoothing pre-processing method, it is possible to detect the frozen and unfrozen of oranges with good accuracy (91.67% of training and 75% of evaluation). Therefore, the FT-IR and Hyperspectral imaging methods are able to detect frozen and unfrozen of oranges (Thomson) but the FT-IR spectroscopic method has a higher accuracy. Moreover, the hyperspectral imaging method provides the best results for the pre-processing of the smoothing type. [ABSTRACT FROM AUTHOR]

Copyright of Innovative Food Technologies (IFT) is the property of Iranian Research Organization for Science & Technology (IROST) and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)