单粒玉米种子水分近红外快速无损测定方法研究

玉米精量播种技术发展对种子质量检测提出了单粒、无损、快速测定等新需求,本研究重点探索了近红外光谱结合化学计量学方法建立单粒玉米种子水分检测模型的可行性。实验收集并测定了110份玉米样本的水分含量,应用傅里叶变换红外光谱仪及单粒测样附件扫描得到样本集近红外光谱,按照3:1随机划分训练集和测试集。首先采用多种光谱预处理方法消除单粒种子采集光谱时由于颗粒形态等引起的噪声干扰,然后分别建立基于PLS线性模型和SVM非线性模型的单粒玉米种子水分近红外检测模型,其中PLS模型测试集的R为0.93,RMSEP为0.86；SVM模型测试集的R达到0.96,RMSEP为0.71。实验结果表明,光谱预处理结合SVM非线性模型可以有效降低单粒玉米种子近红外光谱采集时引入的非线性干扰,有助于提升单粒玉米种子水分近红外快速无损检测实际应用可行性。

Study on Near Infrared Nondestructive Method for Measuring Water Content of Single Corn Seed

The development of precise corn sowing technology has put forward new requirements for single-grain, non-destructive, and rapid determination of seed quality testing. This research focuses on exploring the feasibility of establishing a detection model of single corn seed moisture. The model combines near-infrared spectroscopy with multiple of chemical methods. In the experiment, the moisture content of 110 corn samples was collected and measured. The Fourier transform infrared spectrometer and single-grain sample attachment scan were used to obtain the near-infrared spectrum of the samples, and the training set and test set were randomly divided according to 3: 1. Firstly, a variety of spectral preprocessing methods were used to eliminate noise interference that caused by particle morphology when collecting spectra of single seeds, and then a near-infrared detection model of single corn seeds moisture based on the partial least-square method (PLS) and the model based on Support Vector Machine (SVM) was established. In the model based on PLS evaluation, the correlation coefficients (R) of test set is 0.93 and the Root-Mean-Square Error of Prediction (RMSEP) of test set is 0.86. In the model based on SVM evaluation, the R of test set is 0.96 and the RMSEP of test set is 0.71. The experimental results show that the spectral pre-processing combined with the model based on SVM can reduce the non-linear interference that introduced in the acquisition of near-infrared spectrum of single-grain corn seeds effectively, and improve the practical application of near-infrared non-destructive detection of single-grain corn seeds moisture.