针对目前农业种植选种应用对于带稃壳水稻种子活力分级检测的迫切需求，以及现有通用的糙米检测技术存在的问题，本文提出一种基于近红外超连续激光光谱的水稻种子活力透射光谱检测方法。首先，设计了种子活力近红外吸收光谱检测系统，测量了3种不同年份的带稃壳的水稻种子的近红外吸收光谱，结果显示，水稻种子的活力梯度与近红外吸收光谱的特征吸收峰值相关。然后，采用归一化、二阶导数校正法和正交信号校正相结合优化了种子光谱的预处理算法。最后，建立主成分分析(PCA)模型，对光谱进行降维，确定最佳主成分数目，应用偏最小二乘判别分析(PLS-DA)建立了水稻种子活力分析鉴别模型。分析结果表明，本文设计的透射式吸收光谱检测系统结合PLS-DA判别模型可对不同活力的水稻种子进行分类，校正集和验证集的准确率分别为94.44%和95.92%，筛选后水稻种子的发芽率可达97.17%。研究结果表明，本文提出的基于近红外光谱信息实现水稻种子活力无损分级的方法可行，且具有较高的预测精度。

In view of the urgent need for seed selection technology in agriculture and for grading detection of the vigor of three different years of unpeeled rice seeds, we proposed a new method of detecting the vigor of rice seeds based on near-infrared super-continuous laser spectrum to overcome the significant issues in pre-existing universal brown rice detection technology. Firstly, we design a near-infrared absorption spectroscopy system with which we detect seed viability and measure the NIR spectra of three different years of unpeeled rice seeds. The results show that the activity gradient of the rice seeds is correlated with the characteristic absorption peak of their NIR absorption spectrum. Then, the spectrum of seed is optimized with a pretreatment algorithm of normalization, second derivative correction and orthogonal signal correction. Finally, a Principal Component Analysis (PCA) model is established to reduce the dimension of the spectrum and determine the optimal number of principal components. A Partial Least Squares Discriminant Analysis (PLS-DA) model is established. The analysis results show that the transmission absorption spectrum detection system designed in this paper combined with the PLS-DA discrimination model can classify rice seeds of different vigor with an accuracy of 94.44% and 95.92%. After screening, the germination rate of rice seeds can reach 97.17%. The results show that it is feasible to achieve non-destructive classification of rice seed activity using near-infrared spectroscopy with high accuracy.