中文摘要：

研究了应用高光谱图像技术检测‘霞晖5号’水蜜桃早期冷害的方法。通过分析桃果实采后贮藏期间出汁率的变化，确定冷害发生的初始时间。进而针对早期冷害桃果实的400～1000nm波段的高光谱反射和半透射图像，应用独立主成分分析方法优选出冷害的特征波长，反射条件下为656、674nm和704nm，半透射条件下为640、745nm和811nm，并通过提取每个特征波长下的光谱平均值作为Fisher判别方法建模的特征集，所建水蜜桃早期冷害判别模型的预测准确率分别为83．0％和94．0％。结果表明，高光谱技术可以检测桃果实早期冷害，且半透射照射方式的判别结果优于反射方式。

英文摘要：

Hyperspectral imaging technique was investigated for the detection of chilling injury in ＇Xiahui 5＇ honey peach during the early postharvest storage period. Extractable juice content was measured during the storage to find the time point of occurrance of early chilling injury. Then, hyperspectral reflectance and half-transmission images over the wavelength range of 400-1 000 nm were captured, and independent components analysis （ICA） was implemented on both modes of pictures, respectively. The optimal wavelengths selected by ICA were 656, 674 and 704 nm for reflectance mode, and 640, 745 and 811 nm for half transmittance mode, respectively. Further, spectral average of each characteristic band was chosen as the input of the Fisher discrimination model, an average prediction accuracy of 83.0% and 94.0% for each mode to distinguish between normal and early injured peach were achieved. This research has demonstrated that the hyperspectral imaging technique is feasible for the detection of chilled peach at the early period, and half transmittance mode has a better performance.