| 本文已被：浏览 1255次 下载 1116次 |
码上扫一扫！ |
| 玉米秸秆磷含量近红外漫反射光谱的建模研究 |
|
李冬冬1,王浩瑛1,王蒙1,王铭1,李国梁1,何思洋2,陈绍江1,刘文欣1*
|
|
|
| (1.中国农业大学 农学院/作物杂种优势研究与利用教育部重点实验室/作物遗传改良北京市重点实验室/国家玉米改良中心, 北京 100193;2.中国农业大学 国家能源非粮生物质原料研发中心, 北京 100193) |
|
| 摘要: |
| 利用近红外光谱(NIRS)技术，建立玉米秸秆磷含量的快速定量分析模型，有利于植物养分利用效率和遗传育种研究。本研究选取了来源广泛的200份玉米自交系秸秆样品，结合近红外光谱仪在4 000~10 000 cm-1波段处测量的秸秆样品的光谱数据和钼锑抗比色法测量的样品磷浓度，分别用偏最小二乘(PLS)、最小绝对值收敛和选择算子(LASSO)、支持向量机(SVM)和回归树(RT)的方法，建立了玉米秸秆磷含量的近红外定量分析模型。5折交叉验证结果显示，PLS为最优建模方法，测试集预测值和真实值的相关系数(rtest)为0.80±0.05，训练集的拟合相关系数(rtraining)为0.97±0.03。将PLS与不同的光谱预处理方法结合来优化模型，在11种预处理方法中，归一化以及平滑化的预处理方法效果最好，rtest均为0.80±0.05，与原始光谱数据相比相关性并未显著提高。随着训练集个体的增加，预测准确性不断提高，当训练集和验证集比例为8∶2 时，预测相关系数为0.80，与留一法的预测相关系数(0.81)相当。综上，PLS建立的玉米秸秆磷含量的近红外定量分析模型预测准确性较高，可用于育种或者遗传学研究中磷含量的测定。 |
| 关键词: 近红外光谱 磷含量 偏最小二乘 玉米秸秆 |
| DOI：10.11841/j.issn.1007-4333.2021.08.01 |
| 投稿时间：2020-11-05 |
| 基金项目:国家重点研发计划项目(2016YFD0101201,2018YFD0100201)；中德国际研究培训项目(AMAIZE-P, 328017493/GRK 2366) |
|
| Near-infrared reflectance spectroscopy to analysis phosphorus concentration in maize straw |
|
LI Dongdong1,WANG Haoying1,WANG Meng1,WANG Ming1,LI Guoliang1,HE Siyang2,CHEN Shaojiang1,LIU Wenxin1*
|
| (1.Key Laboratory of Crop Heterosis and Utilization, the Ministry of Education/Key Laboratory of Crop Genetic Improvement, Beijing Municipality/National Maize Improvement Center/College of Agronomy and Biotechnology, China Agricultural University, Beijing 100193, China;2.National Research and Development Center of Energy Non-grain Biomass Raw Materials, China Agricultural University, Beijing 100193, China) |
| Abstract: |
| Research in use of near-infrared spectroscopy(NIRS)to establish a fast-quantitative analysis model of measuring maize straw phosphorus concentration will benefit the genetic study of plant nutrient utilization efficiency and breeding. A total of 200 maize straw samples were randomly selected. These samples were measured by both NIRS(at a band width of 4 000-10 000 cm-1)and by chemical method(molybdenum antimony colorimetric method). Four modeling methods including partial least squares(PLS), least absolute shrinkage and selection operator(LASSO), support vector machine(SVM), and regression tree(RT)were then used to establish NIRS prediction models for phosphorus concentration in maize straw. The fivefold cross-validation results showed that PLS was the most optimal modeling method among the four approaches. For PLS, the correlation coefficient(rtest)between predicted and measured phosphorus concentration values was 0. 80±0. 05 in the test set; and the correlation coefficient(rtraining)was 0. 97±0. 03 in the training set. Furthermore, the effects of eleven pre-processing methods based on PLS were evaluated. The results showed that the normalization and smoothing performed the best, with rtest of 0. 80±0. 05. However, no significant difference was observed compared with the original data. The results showed that the prediction accuracy was improved with the increase of the training set size. When the ratio of training set to test set was 8∶2, the prediction accuracy was 0. 80, which was equivalent to the prediction accuracy of Leave-One-Out method(0. 81). To sum up, the NIRS quantitative model of maize straw phosphorus concentration by PLS has a high prediction accuracy and can be used for the determination of phosphorus concentration in breeding and genetic studies. |
| Key words: near-infrared reflectance spectroscopy phosphorus concentration partial least square maize straw |
