Apples are highly susceptible to contamination from environmental factors and fungi during storage and logistics, leading to spoilage and economic losses. The study proposes a method for batch monitoring of apple quality and real-time spoilage warning system, based on multi-sensor fusion technology (MSFT), deep learning (DL), and variable selection algorithms (VSA). Volatile gases from apples infected with Colletotrichum acutatum (C.a), Botrytis cinerea (B.c), and Botryosphaeria sarmentorum (B.s) were collected using gas sensors. Key feature variables were optimized through VSA, and a prediction model for spoilage was developed using chemometrics and DL. Models based on Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) demonstrated strong predictive performance, with correlation coefficients (R) exceeding 0.88. Optimal performance was observed in the Synergy Interval-Gated Recurrent Unit (Si-GRU) model for B.s, with R of 0.9473 for calibration (R c ) and 0.9430 for prediction (R p ). A vehicle-mounted terminal using MSFT was developed to acquire temperature, humidity, vibration, and gas data, and transmitted this information in real time to a remote monitoring and warning platform. The Si-GRU model was identified as optimal for constructing a multi-environmental factor spoilage warning model, achieving R above 0.98, indicating robustness and accuracy. The optimal model was implemented on the platform, providing visual displays and graded alerts for apple spoilage. Results indicated that key quality indicators were effectively captured by MSFT, and primary spoilage factors were identified by VSA, enabling accurate DL-based monitoring and real-time warning of apple spoilage.
翻译：苹果在储存和运输过程中极易受环境因素和真菌污染，导致腐败和经济损失。本研究提出了一种基于多传感器融合技术、深度学习和变量选择算法的苹果品质批次监测与实时腐败预警系统。使用气体传感器收集感染了炭疽病菌、灰霉病菌和贝仑格葡萄座腔菌的苹果释放的挥发性气体。通过变量选择算法优化关键特征变量，并利用化学计量学和深度学习建立腐败预测模型。基于长短期记忆网络和门控循环单元的模型表现出较强的预测性能，相关系数超过0.88。其中，针对贝仑格葡萄座腔菌的协同区间-门控循环单元模型效果最佳，其校正相关系数和预测相关系数分别为0.9473和0.9430。开发了一种采用多传感器融合技术的车载终端，用于采集温度、湿度、振动和气体数据，并将这些信息实时传输到远程监测预警平台。Si-GRU模型被确定为构建多环境因子腐败预警模型的最佳模型，其相关系数超过0.98，表明模型稳健且准确。该最优模型在平台上实现，为苹果腐败提供了可视化显示和分级警报。结果表明，多传感器融合技术能有效捕捉关键品质指标，变量选择算法能识别主要腐败因素，从而实现了基于深度学习的苹果腐败准确监测与实时预警。

The authenticity of wine is a significant concern in the global market due to widespread fraudulent activities involving wine origins. In this study, the inorganic and organic constituent profiles of 102 wine samples from five different growing regions were analyzed. Low- and mid-level data fusion coupled with a multivariate analysis approach was applied to three platforms of mass spectrometry data sets using Inductively Coupled Plasma-Mass Spectrometry (ICP-MS), Isotopic Ratio Mass Spectrometry (IRMS), and Ultra-high Performance Liquid Chromatography coupled with Ion Mobility Quadrupole Time-Of-Flight Mass Spectrometry (UPLC/IM-QTOF-MS) to determine the correct classification of wine origin. Significant difference of organic and inorganic compounds from five regions were selected for low- and mid-level data fusion. In the low-level data fusion, Partial Least Squares Discriminant Analysis (PLS-DA) under the positive ionization modes (ESI+) achieved an optimal prediction rate of 74.4%. By extracting features from low-level data fusion for mid-level data fusion, linear discriminant and nonlinear models were established, resulting in an optimal accuracy of 98.7%. The mid-level data fusion multivariate analysis strategy significantly enhanced the accuracy of identifying the geographical origin of wine and presents a new approach for identifying origins in the origin of other foods.
翻译：由于涉及葡萄酒产地的欺诈活动普遍存在，葡萄酒的真实性是全球市场的一个重大问题。本研究分析了来自五个不同产区的102个葡萄酒样品的无机和有机成分谱。采用低层级和中层级数据融合策略，结合多元分析方法，处理了来自电感耦合等离子体质谱、同位素比值质谱和超高效液相色谱-离子淌度四极杆飞行时间质谱的三个质谱数据集，以确定葡萄酒产地的正确分类。筛选出五个产区间存在显著差异的有机和无机化合物进行低层级和中层级数据融合。在低层级数据融合中，基于正离子模式的偏最小二乘判别分析取得了74.4%的最佳预测率。通过从低层级融合数据

The diversity of botanical origins may influence the composition of honey and thus its recognition as a functional and healthy food. This study examined the standard physicochemical properties, bioactive components and antioxidant activity of Kosovan honeys according to their floral source (monofloral, blossom, acacia, and mountain blossom honey). Then the Kohonen Neural Network (KNN), which transforms complex multivariate data into two-dimensional space, and Principal Component Analysis (PCA) were used to identify and group botanical origin of honey samples based on their component features. Physicochemical characteristics, total phenolic content, and antioxidant activity varied significantly between the individual distinct varieties of honeys. Statistical analysis showed the usefulness of KNN and PCA for dimensionality reduction and detecting the structure and general regularities in the values of variables describing the tested honeys of the same botanical origin. KNNs have proven to be a particularly effective data mining tool, enabling the detection of subtle differences and clearer separation of clusters occurring in honey samples. The developed KNN model revealed proximity between the AC and MBL clusters, as well as between the MF and BL clusters, indicating similarity of their features. The arrangement of honey groups on the matrix map also suggested that the properties of AC and MBL honeys were significantly different from those of MF and BL honeys. The research showed that both methods used could be used as additional statistical tools supporting the recognition of the type of honey according to its chemical composition, mineral content, bioactive components and the antioxidant activity of honey as a functional food.
翻译：植物来源的多样性可能影响蜂蜜的组成，进而影响其作为功能性和健康食品的认可度。本研究根据花源（单花蜜、百花蜜、洋槐蜜和高山百花蜜）检测了科索沃蜂蜜的标准理化性质、生物活性成分和抗氧化活性。然后，使用将复杂多元数据转换到二维空间的Kohonen神经网络（KNN）和主成分分析（PCA），根据成分特征对蜂蜜样品的植物来源进行识别和分组。不同品种蜂蜜的理化特性、总酚含量和抗氧化活性存在显著差异。统计分析表明，KNN和PCA在降维以及检测描述相同植物来源蜂蜜的变量值结构和一般规律方面非常有用。KNN被证明是一种特别有效的数据挖掘工具，能够检测蜂蜜样本中细微的差异并实现更清晰的聚类分离。所开发的KNN模型揭示了AC和MBL聚类之间以及MF和BL聚类之间的接近性，表明它们的特征相似。蜂蜜群落在矩阵图上的排列也表明，AC和MBL蜂蜜的特性与MF和BL蜂蜜的特性有显著不同。研究表明，所使用的两种方法均可作为额外的统计工具，支持根据蜂蜜的化学成分、矿物质含量、生物活性成分以及作为功能性食品的抗氧化活性来识别蜂蜜类型。

Chilling injury in coriander (Coriandrum sativum L.) occurs easily during transportation at low temperatures and cold storage to maintain freshness, leading to quality deterioration. Such chilling injury is difficult to notice at the early stage, thus rapid and accurate detecting is necessary. In this study, the application of short-wave infrared hyperspectral imaging (SWIR-HSI) combined with machine learning (ML) techniques was demonstrated in quality control of coriander. Selection of the relevant wavelength in SWIR region was conducted prior to investigation of ML algorithms for effective classification of chilling injury. The early stage of chilling injury was explored with SWIR-HSI in comparison with conventional measurement of moisture, total chlorophyll and color. Notably, early chilling injury was detected after 7 days of cold storage with SWIR-HSI, whereas it required 14 days for observation by conventional measurements. In addition, different profiles of volatile organic compounds from different cultivation batches were recognized by the current application of SWIR-HSI, which was in agreement with results from gas chromatography-mass spectrometry. These findings suggested the potential of SWIR-HSI in combination with ML, as a rapid and accurate technique for determining quality of coriander. Among the ML algorithms tested, i.e., K-Nearest Neighbors, Support Vector Machine, Random Forest Classifier, Multilayer Perceptron, and Convolutional Neural Network (CNN), CNN coupled with CNN-based wavelength selection yielded the best performance in classifying the early chilling injury with 94% accuracy.
翻译：香菜在低温运输和冷藏保鲜过程中极易发生冷害，导致品质劣变。这种冷害在早期阶段难以察觉，因此需要快速准确的检测方法。本研究展示了短波红外高光谱成像技术结合机器学习在香菜质量控制中的应用。在评估机器学习算法对冷害的有效分类之前，先进行了SWIR区域相关波长的选择。通过与水分、总叶绿素和颜色等传统测量方法进行比较，探索了SWIR-HSI对早期冷害的检测能力。值得注意的是，SWIR-HSI在冷藏7天后即能检测到早期冷害，而传统测量方法则需要14天才能观察到。此外，通过SWIR-HSI应用识别了不同种植批次香菜的挥发性有机化合物谱差异，结果与气相色谱-质谱分析一致。这些发现表明了SWIR-HSI结合机器学习作为快速准确判定香菜品质技术的潜力。在所测试的K近邻、支持向量机、随机森林分类器、多层感知机和卷积神经网络等机器学习算法中，CNN结合基于CNN的波长选择在早期冷害分类中表现最佳，准确率达到94%。

¹H NMR fingerprinting of virgin olive oils (VOOs) and a collection of binary classification models arranged in a decision tree are presented as a stepwise strategy to determine the geographical origin of a VOO at four levels, i.e. provenance from an EU member state or outside the EU, country and region of origin, and compliance with a geographical indication scheme. This approach supports current EU regulation that makes labelling of the geographical origin mandatory for olive oil. Currently, official methods for its control are still lacking. Partial least squares discriminant analysis (PLS-DA) and random forest for classification afforded robust and stable binary classification models to verify the geographical origin of VOOs; however, the former outperformed the latter in terms of accuracy and robustness. The prediction abilities of the best binary PLS-DA model for each case study were between 80% and 100% for both classes in cross-validation and in external validation. The satisfactory results achieved for the verification of the geographical origin of VOOs, together with those of our previous studies on the discrimination of olive oil categories, the detection of olive oils blended with vegetable oils, and the determination of the stability, freshness, storage time and conditions, and olive oil best− before date, confirm that a single ¹H NMR analysis of an olive oil sample can provide useful information to control several EU regulations related to olive oil marketing standards.
翻译：本文提出了一种基于初榨橄榄油(VOO)的¹H核磁共振指纹图谱和以决策树结构组织的一系列二元分类模型的分步策略，用于从四个层级鉴定橄榄油的地理来源：即来自欧盟成员国或非欧盟国家、具体国家和产区的来源，以及是否符合地理标志保护。该方法旨在支持当前规定橄榄油标签必须注明地理来源的欧盟法规。目前，官方尚缺乏相应的监控方法。偏最小二乘判别分析(PLS-DA)和随机森林分类算法均能提供稳健可靠的二元分类模型来验证橄榄油的地理来源；但前者在准确性和稳健性方面优于后者。针对每个案例构建的最佳二元PLS-DA模型，其预测能力（包括交叉验证和外部验证）在两类上的准确率均介于80%至100%之间。验证橄榄油地理来源所获得的满意结果，连同我们先前关于橄榄油类别区分、检出橄榄油中掺混植物油、以及确定稳定性、新鲜度、储存时间和条件与最佳食用日期等方面的研究，证实了对单个橄榄油样本进行一次¹H NMR分析，即可为监管与橄榄油市场标准相关的多项欧盟法规提供有用信息。