糖度和硬度作为水蜜桃的两个重要指标, 决定其内部品质。 在运输或售卖期间, 水蜜桃果内水分流失, 表面开始松软进而腐烂, 内部品质发生变化。 研究旨在探讨可见/近红外光谱预测水蜜桃不同贮藏期糖度和硬度的可行性, 进一步预测水蜜桃的最佳贮藏期。 采用漫透射和漫反射方式采集4个贮藏阶段的水蜜桃光谱, 并测量糖度和硬度。 分析了4个阶段水蜜桃的平均光谱, 光谱强度随着贮藏天数增加而不断提高, 且在650~680 nm区域内受果皮颜色及色素的变化产生波峰偏移。 同时, 分析了糖度和硬度的变化, 糖度在贮藏期间逐渐提高, 硬度在贮藏期间快速下降, 最终糖度增加了3.31%, 硬度下降了58.8%。 采用多元散射校正、 S-G卷积平滑、 归一化处理及基线校正等预处理方法来减少噪声和误差对光谱的影响, 并使用无信息变量消除(UVE)和连续投影算法(SPA)筛选特征波长, 最后利用偏最小二乘回归(PLS)分别建立糖度和硬度的预测模型。 分析糖度、 硬度的PLS回归系数与平均光谱的波形发现, 糖度的高回归系数分布在光谱多处, 而硬度的该系数均在波峰波谷附近。 SPA和UVE筛选的特征波长建立的糖度模型效果不佳, 而硬度模型效果良好。 结果表明, 漫透射和漫反射检测方式下, 糖度的最佳预测相关系数(R_p)及预测均方根误差(RMSEP)分别为0. 886, 0.727和0.820, 1.003, 预处理方法分别是多元散射校正、 平滑窗口宽度为3的S-G卷积平滑。 此外, 漫透射建立的硬度SPA-PLS模型, 选用15个光谱变量, 得到的R_p和RMSEP为0.798和0.976; 而漫反射建立的UVE-PLS模型, 选用113个光谱变量, 得到的R_p和RMSEP为0.841和0.829。 可以看出, 漫透射方式预测水蜜桃贮藏期间的糖度更佳, 而漫反射预测硬度更佳。 利用可见/近红外光谱所建立的糖度和硬度预测模型, 能够可靠地预测水蜜桃贮藏期内糖度和硬度的变化, 对指导采摘、 售卖时间和减少腐烂具有一定的参考价值。

Soluble solid content (SSC) and firmness are two important indexes of peach, which determine its internal quality. However, the water in the peach fruit is lost, the surface begins to soften and rot, and the internal quality changes during transportation or sale. This paper aims to investigate the feasibility of visible/near-infrared spectroscopy(VIS-NIR)in predicting SSC and firmness of peach during different storage days and to predict the optimal storage period of peaches further. The spectrum of peach in 4 storage stages was collected by diffuse transmittance and reflectance, and the sugar and hardness were measured. The mean spectrum of peach in four stages was analyzed. The spectral intensity increased with the storage days, and the peak shift was caused by the changes in the color and pigment of the peel in the region of 650~680 nm. Meanwhile, the changes in SSC and firmness were analyzed. The SSC gradually increased during storage, while the firmness rapidly decreased during storage. Finally, the SSC increased by 3.31% and the firmness decreased by 58.8%. Pretreatment methods such as multivariate scattering correction(MSC), S-G smoothing, normalization and baseline are used to reduce the impact of noise and errors in the spectrum, and uninformative variable elimination (UVE) and successive projections algorithm (SPA) is used to select characteristic wavelengths, then the partial least squares regression(PLS) is used to establish prediction models for SSC and firmness. Analyzing the PLS regression coefficient of SSC and firmness with the mean spectrum, it is found that SSC has many high regression coefficient bands, and the high regression coefficient of firmness is near the peaks and troughs. Therefore, the SSC model established by the characteristic wavelength obtained by SPA and UVE is not good, while the firmness model is good. The results show that the best prediction correlation coefficient (R_p) and root mean square error of prediction (RMSEP) of SSC under the diffuse transmittance and reflectance detection methods are 0.886, 0.727, 0.820, 1.003, respectively. The pretreatment methods are MSC and S-G smoothing with 3 smoothing window width, respectively. In addition, the SPA-PLS model of firmness established by diffuse transmittance uses 15 spectral variables to obtain R_p and RMSEP of 0.798 and 0.976. The UVE-PLS model established by the diffuse reflectance uses 113 spectral variables to obtain R_p and RMSEP of 0.841 and 0.829. It can be seen that the diffuse transmittance method predicts SSC better, and the diffuse reflectance predicts the firmness better during peach storage. The SSC and firmness prediction models established by VIS-NIR can reliably predict the changes of SSC and firmness during the storage of peaches and have certain reference value to guide picking and selling time and reduce decay.