葡萄中可溶性固形物是评价葡萄成熟度的重要指标，本文探究了基于可见/近红外光谱技术对多个品种葡萄（红提、巨峰、辽峰）可溶性固形物（Soluble Solid Content，SSC）含量进行定量分析。分别采集了三个葡萄品种在550-960nm波长范围内的透射光谱数据，采用Savitzky-Golay卷积平滑（S-G）、标准正态变换（Standard Normal Variate, SNV）、小波变换（WT）、一阶求导+S-G卷积平滑组合（1stDer+S-G）预处理方法，对比分析出最适合各个品种的预处理方法；然后在最佳的预处理方法下采用连续投影算法(SPA)、竞争性自适应重加权（CARS）对光谱进行特征波长选择；结合化学计量学方法分别建立多品种与单一品种的偏最小二乘回归（PLSR）、BP神经网络SSC含量无损预测模型。结果表明，基于BP-SPA建立的SSC含量模型最优，多个品种通用SSC含量预测模型的预测集相关系数（Rp2）为0.85，表明基于可见/近红外光谱技术对多个葡萄品种SSC含量无损检测是可行的。

The Soluble solid content in grapes is an important indicator for evaluating grape ripeness, and this paper explores the quantitative analysis of soluble solid content (SSC) content in several varieties of grapes (Hongti, Jufeng, and Liaofeng) based on visible/near-infrared (NIR) spectroscopic techniques. The transmission spectra of three grape varieties in the wavelength range of 550-960 nm were collected separately, and Savitzky-Golay convolutional smoothing (S-G), standard normal variate (SNV), wavelet transform (WT), and the combination of first-order derivation + S-G convolutional smoothing (1stDer+S-G) were used to analyze the soluble solids content of the grapes. Preprocessing methods, and compare and analyze the most suitable preprocessing methods for each variety; then under the optimal preprocessing methods, we used the continuous projection algorithm (SPA) and competitive adaptive reweighting (CARS) to select the characteristic wavelengths of the spectra; and combined with chemometrics methods to establish the partial least squares regression (PLSR) for multi-species and single-species, and the lossless prediction model for the content of the SSC of the BP neural network, respectively. The results showed that the SSC content model based on BP-SPA was optimal, and the prediction set correlation coefficient (Rp2) of the generalized SSC content prediction model for multiple varieties was 0.85, which indicated that the non-destructive detection of SSC content in multiple grape varieties based on visible/near-infrared spectroscopy was feasible.