现阶段水稻种子发芽率测试仍然按照传统的农作物种子发芽技术规定进行发芽试验, 此方法存在试验周期长、 成本高、 专业性要求高等缺点, 本研究提出一种基于近红外光谱技术的快速、 无损测试杂交水稻种子发芽率的新方法。 采用人工老化方法在温度45 ℃、 湿度100%的条件下分别老化处理2个品种杂交水稻种子0, 24, 48, 72, 96, 120, 144 h; 用近红外光谱仪分别采集2个品种不同老化时间段杂交水稻种子光谱数据共280份, 随机分成校正集（168份）和检验集（112份）; 测试不同老化时间段的水稻种子发芽率; 以偏最小二乘算法（PLS）建立了回归模型, 分析不同光谱波段和比较不同光谱预处理方法对模型精度的影响。 2个品种的水稻种子光谱数据采用全波段和标准化＋正交信号校正预处理时模型最优, 模型校正集决定系数（RC）与验证集相关系数（RP）分别为0.965和0.931, 校正标准误差（SEC）与预测标准误差（SEP）分别为1.929和2.899, 验证集预测值与真实值之间的相对误差在4.2%以内。 研究结果表明利用近红外光谱分析技术进行杂交水稻种子发芽率的快速无损检测是可行的。

Germination rate of rice seeds was measured according to technical stipulation of germination testing for agricultural crop seeds at present. There existed many faults for this technical stipulation such as long experimental period, more costing and higher professional requirement. A rapid and non-invasive method was put forward to measure the germination rate of hybrid rice seeds based on near-infrared reflectance spectroscopy. Two varieties of hybrid rice seeds were aged artificially at temperature 45 ℃ and humidity 100% condition for 0, 24, 48, 72, 96, 120 and 144 h. Spectral data of 280 samples for 2 varieties of hybrid rice seeds with different aging time were acquired individually by near-infrared spectra analyzer. Spectral data of 280 samples for 2 varieties of hybrid rice seeds were randomly divided into calibration set (168 samples) and prediction set (112 samples). Gormination rate of rice seed with different aging time was tested. Regression model was established by using partial least squares (PLS). The effect of the different spectral bands on the accuracy of models was analyzed and the effect of the different spectral preprocessing methods on the accuracy of models was also compared. Optimal model was achieved under the whole bands and by using standardization and orthogonal signal correction (OSC) preprocessing algorithms with CM2000 software for spectral data of 2 varieties of hybrid rice seeds, the coefficient of determination of the calibration set (RC) and that of the prediction set (RP) were 0.965 and 0.931 individually, standard error of calibration set (SEC) and that of prediction set (SEP) were 1.929 and 2.899 respectively. Relative error between tested value and predicted value for prediction set of rice seeds is below 4.2%. The experimental results show that it is feasible that rice germination rate is detected rapidly and nondestructively by using the near-infrared spectroscopy analysis technology.