种子活力对于农业发展至关重要, 而甜玉米种子普遍存在活力较低且不耐贮藏的问题。 因此, 及时准确地对甜玉米种子活力进行检测尤为重要。 电导率测定法作为一种传统的种子活力检测方法, 存在对种子有一定破坏性、 耗时较长、 重复性不佳等缺点。 针这些问题, 尝试利用可见-近红外(VIS-NIR)高光谱成像系统结合化学计量学算法建立甜玉米种子电导率快速、 无损且精确的检测方法。 以高温高湿老化的绿色超人甜玉米种子为试验材料, 先通过可见-近红外高光谱成像系统采集种子的高光谱图像和进行电导率测定试验, 随后对高光谱图像进行黑白板校正、 提取感兴趣区域, 获取光谱反射率数据。 利用多种预处理方法分别为标准正态变量变换(SNV)、 二阶导(SD)、 一阶导(FD)、 和多元散射校正(MSC)建立甜玉米种子电导率的偏最小二乘回归(PLSR)模型, 比较分析并筛选出最适预处理方法。 再通过连续投影算法(SPA)及遗传算法(GA)对MSC预处理后的高光谱波段进行筛选提取, 基于选出的特征波段建立PLSR模型, 并与全波段(Full)PLSR模型进行对比分析, 得到与甜玉米种子电导率相关性最高的高光谱波段组合, 最终确立一种能够预测甜玉米种子电导率的方法体系。 实验结果显示: 不同预处理方法(SNV, FD, SD和MSC)建立的PLSR模型性能有所差异, 其中MSC-PLSR模型的表现最优秀, 其校正决定系数和预测决定系数分别为0.983和0.974, 相应的校正均方根误差和预测均方根误差分别为0.165和0.226。 进一步分析MSC-Full-PLSR, MSC-SPA-PLSR和MSC-GA-PLSR模型, 发现GA能够将全光谱的853个波段压缩至25个有效波段, 所建立的MSC-GA-PLSR模型仍表现优秀, 其校正决定系数和预测决定系数分别为0.976和0.973, 相应的校正均方根误差和预测均方根误差分别为0.194和0.212。 实验结果表明: 基于可见-近红外(VIS-NIR)高光谱成像系统结合化学计量学算法实现对甜玉米种子电导率的预测存在一定的可行性。 该研究为甜玉米种子电导率的快速、 无损且精确的检测提供一定的理论支持。

The vigor of seeds plays a vital role to the agricultural development. But the low vigor and storage-tolerance seeds are common problems for sweet corn. Therefore, it has a certain practical significance to detect the sweet corn seed vigor accurately and timely. Electrical conductivity test is a traditional method of determining the vigor ofseeds. However, it is a labor-intensive, time-consuming, and destructive process, which is subject to human error. Given that, this study investigated the possibility of using visible and near-infrared (VIS/NIR) hyperspectral imaging (HSI) technique to detect the electrical conductivity of sweet corn seeds. Sweet cornseeds treated by high temperature and high humidity aging were prepared as experimental materials. The visible and near-infrared hyperspectral imaging acquisition system (400~1 000 nm) was constructed to acquire the hyperspectral images of the sweet corn seeds. After HSI spectra collection, electrical conductivity tests were conducted in sweet corn seeds. The average reflectance data of the region of interest were extracted for spectral characteristics analysis. Then different pre-processing algorithms including standard normal variate (SNV), first derivative (FD), second derivative(SD), multiplicative scatter correction (MSC) were conducted to build partial least squares regression (PLSR) models of the conductivity. Lastly, the hyperspectral effective wavelengths related to conductivity of sweet corn seeds were extracted by SPA and GA for PLSR models. The results showed that the best pre-processing algorithm was MSC method. The SPA was not performing as well as GA which selected only 25 characteristic wavebands from the all 853spectral wavebands. The PLSR model built by using MSC and GA exhibited the optimal performance with correlation coefficient of 0.976 and 0.973 for calibration set and prediction set, respectively, and root mean squared error for calibration and prediction were 0.194 and 0.212. The results indicated that combining the visible and near-infrared hyperspectral imaging technique with MSC-GA-PLSR can be used as a feasible and reliable method for the determination of conductivity in sweet corn seeds. The result can provide a theoretical foundation for rapid detection of seed conductivity using spectral information.