快速检测活体水果内部品质对于确定水果最佳采摘时机和果园信息化管理具有重要意义。 以南方棚栽葡萄为研究对象，应用光谱技术对处于生长期的四个葡萄品种的可溶性固体含量(SSC)进行现场测试。 分别采用偏最小二乘法(PLS)回归、潜变量人工神经网络(LV-ANN)和潜变量支持向量机(LV-SVM) 三种方法为光谱建模集建立了SSC校正模型。用验证集对模型的预测性能进行了评价。与PLS和LV-ANN模型相比，LV-SVM模型的预测性能最佳。 实验结果表明，将光谱技术与LV-SVM建模法相结合适用于果园葡萄活体可溶性固体含量无损检测。

The fast detection of inner quality of living fruit is of importance to the selection of optimal harvest time and to the information management of an orchard. The trellised grapes in the southern part of our country are used as the research object. The soluble solid content (SSC) of four kinds of grapes in growth is detected by using a visible and near infrared spectrophotometer on site. The SSC correction models are established by using Partial Least Square regression (PLS) , Latent Variable and Artificial Neural Network (LV-ANN) and Latent Variable and Support Vector Machine (LV-SVM) respectively. The prediction performance of these models is evaluated by using a validation set. Compared with the PLS and LV-ANN models, the LV-SVM model has the best prediction performance. The experimental result shows that the combination of spectroscopy with the LV-SVM modeling is suitable for the nondestructive SSC detection of living grapes in an orchard.