Near-infrared (NIR) spectroscopy is a useful technique for the non-destructive analysis of fruit quality. The key quality parameters of table grapes ( Vitis vinifera ) that affect consumer preference are the soluble solids content (SSC), pH, firmness, and seedlessness. This research focused on using NIR spectroscopy for assessing the quality of ‘Kyoho’ table grapes, as a non-destructive analysis under laboratory and field conditions. NIR spectra for each sample were acquired in the wavelength range of 400–1000 nm, using a visible/NIR spectrometer with fibre optics in the interactance mode. Partial least-square regression was used to calibrate the NIR spectral data with all the measured properties of table grapes. The best prediction model for firmness was the Savitzky–Golay first derivative (SGD1) with a coefficient of determination (R prediction 2 ) of 0.7427 in the laboratory, and 0.7804 in the field. The R prediction 2 values for pH in the laboratory and the field was 0.6276 using multiplicative scatter correction (MSC), and 0.7676 using SGD1, respectively. These values were similar to the R prediction 2 values of SSC, which were 0.6926 using MSC, and 0.8052 using the Savitzky–Golay second derivative, respectively. In both analyses the R 2 of the calibration model was between 0.6944 and 0.8877. The partial least-square discriminant analysis was used to classify the percentage of seedlessness, which was 93.10% in the laboratory using SGD1 or MSC, and 79.31% in the field using MSC. Therefore, NIR spectroscopy is an efficient non-destructive technique for rapidly analysing Japanese table grape qualities in laboratory and field settings.

中文翻译：

---

使用近红外光谱法无损分析日本鲜食葡萄品质

近红外 (NIR) 光谱是一种用于水果品质无损分析的有用技术。影响消费者偏好的鲜食葡萄 (Vitis vinifera) 的关键质量参数是可溶性固形物含量 (SSC)、pH、硬度和无籽。这项研究的重点是使用 NIR 光谱评估“巨峰”鲜食葡萄的质量，在实验室和现场条件下进行无损分析。每个样品的 NIR 光谱在 400-1000 nm 的波长范围内获得，使用可见光/NIR 光谱仪在交互模式下使用光纤。偏最小二乘回归用于校准具有鲜食葡萄所有测量特性的 NIR 光谱数据。硬度的最佳预测模型是 Savitzky-Golay 一阶导数 (SGD1)，其决定系数 (R 预测 2 ) 在实验室中为 0.7427，在现场为 0.7804。实验室和野外 pH 值的 R 预测 2 值使用乘法散射校正 (MSC) 分别为 0.6276，使用 SGD1 时分别为 0.7676。这些值与 SSC 的 R 预测 2 值相似，使用 MSC 时分别为 0.6926，使用 Savitzky-Golay 二阶导数时分别为 0.8052。在这两种分析中，校准模型的 R 2 都在 0.6944 和 0.8877 之间。采用偏最小二乘判别分析对无籽百分比进行分类，在实验室使用SGD1或MSC时为93.10%，在田间使用MSC时为79.31%。所以，