为有效筛选F1代玉米杂交种的优势种, 提出了一种基于近红外光谱技术和二维相关谱技术相结合的种子快速筛选方法, 以内蒙古奈曼旗当地的推广品种京科958(JK958)作为对照, 测定了26个玉米品系的近红外光谱, 并对其进行聚类分析。 当类间距=10时, 26个品系被聚成3类。 第一类是与对照品相近的品系: JK958(对照品种), 26, 14, 489, 263和320等； 第二类是9, 542, 16, 121以及57品系； 第三类是与对照品较远的品系: 317, 582, 284, 264以及157。 一般, 若两个品系差别越小, 其自相关谱强度就越小, 反之成立。 若两个品系完全相同, 理想情况下同步二维相关谱上不会出现任何特征信息, 因此可通过自相关谱强度来判别两个品系的相似程度。 根据聚类分析结果, 选出了品系14, 26, 9和157进行二维相关性分析。 其中, 品系14和26是在近红外光谱聚类分析中与对照品种JK958相距较近的品系, 而9和157则与JK958相距较远。 品系14和JK958自相关强度的范围在0.000 0～0.000 2 a.u.； 品系26和JK958自相关强度的范围为: 0.000 00～0.000 10 a.u.； 品系9与JK958的自相关强度范围: 0.000 0～0.001 6 a.u.； 品系157与JK958的自相关强度约为0.000 4～0.002 0 a.u.； 从自相关强度整体来看四个品系与JK958的自相关强度由大到小的排列顺序为157>9>14>26, 表明: 品系26与对照品系JK958最相似。 为了验证上述筛选品系方法的有效性, 将以田间实地测量的玉米15个主要农艺性状数据, 进行聚类分析, 并与近红外光谱和二维相关谱快速筛选的结果进行对照。 研究发现: 农艺性状的聚类分析结果与近红外光谱聚类结果存在交叉(品系320, 26, 24, 147, 109和263)。 自相关强度为10-5数量级的品系26与JK958距离较近, 被聚到一类； 而自相关强度为10-4数量级的品系14, 9和157被聚为一类。 农艺性状的聚类结果证实了近红外光谱和二维相关谱技术快速筛选玉米品系的有效性。 研究结果表明: 近红外光谱+二维相关谱技术对玉米品系进行初步快速筛选是可行的和有效的。

In order to screen the dominant species of hybrid corn F1 generation effectively, a fast seed screening method was proposed based on near infrared spectroscopy and two-dimensional (2D) correlation spectroscopy. The jingke 958 (JK958) which is a variety of Naiman banner in Inner Mongolia local promotion as a control sample, the near-infrared (NIR) spectra of 26 maize lines were collected and clustering analysis was carried out. When the class spacing was equal to 10, 26 samples were clustered into 3 groups in terms of the results of cluster analysis. The first group is close to the control sample: JK958, 26, 14, 489, 263, 320 etc. The second group is 9, 542, 16, 121 and 57; the third group is far from the control sample: 317, 582, 284, 264 and 157. In general, if the difference between the two samples is small, the autocorrelation intensity is small. If the two samples are identical, synchronous 2D correlation spectrum cannot represent characteristic information under ideal conditions. Therefore, the similarity between the two samples can be judged by the autocorrelation intensity. According to the results of cluster analysis, lines 14, 26, 9 and 157 were selected for 2D correlation analysis. Among them, lines 14 and 26 are close to JK958 (contrast sample) in NIR cluster analysis, while the 9 and 157 with JK958 are far apart. The autocorrelation intensity is in the range of 0.000 0～0.000 2 a.u. for sample 14, 0.000 00～0.000 10 a.u. for sample 26, 0.000 0～0.001 6 a.u. for sample 9, 0.000 4～0.002 0 a.u. for sample 157. As a whole, the order of autocorrelation intensity between four samples and JK958 sample is 157>9>14>26, which shows that sample 26 is the most similar to that of control sample JK958. In order to verify the validity of the above screening methods, cluster analysis was carried out based on 16 Main Agronomic Traits of maize measured in the field, and compared with the results of rapid screening by near infrared spectroscopy and 2D correlation spectroscopy. It can be found that there is a cross between the agronomic characters of clustering analysis and the near infrared spectral clustering (line 320, 26, 24, 147, 109 and 263). Samples 26 with the 10-5 autocorrelation intensity was close to JK958, and were clustered into one group, while samples 14, 9 and 157 with the 10-4 autocorrelation intensity were clustered into one group. The clustering results of agronomic traits confirmed the validity of NIR spectroscopy for rapid screening of maize lines. The results showed that NIR spectroscopy combined with 2D correlation spectroscopy was feasible and effective for rapid screening of maize lines.