首页 | 本学科首页   官方微博 | 高级检索  
     

多元散射校正和逐步回归法建立黑土有机碳近红外光谱定量模型
引用本文:申艳,张晓平,梁爱珍,时秀焕,范如芹,杨学明. 多元散射校正和逐步回归法建立黑土有机碳近红外光谱定量模型[J]. 农业系统科学与综合研究, 2010, 26(2): 174-180
作者姓名:申艳  张晓平  梁爱珍  时秀焕  范如芹  杨学明
作者单位:1. 中国科学院,东北地理与农业生态研究所,吉林,长春,130012;中国科学院,研究生院,北京,100049
2. 中国科学院,东北地理与农业生态研究所,吉林,长春,130012
基金项目:国家自然科学基金,吉林省科技发展计划项目,国家科技支撑项目 
摘    要:采用全谱建立近红外校正模型,计算工作量大且冗余信息多。通过特定方法筛选特征波长有可能得到更好的定量校正模型。逐步回归法在波长选取方面起到了重要的作用。多元散射校正技术(Multiple scatter correction,MSC)可以有效剔除由样品颗粒大小、装填密度、湿度等不同引起的散射影响,有效提高光谱的信噪比。以我国东北黑土为研究对象,采集了136个土壤样品3699cm^-1-12000cm^-1范围的近红外光谱,利用多元散射校正技术对近红外原始光谱数据进行预处理,评价MSC的去噪效果,并利用多元线性回归法建立了土壤有机碳含量与校正光谱优选波长点处吸光度之间的关系模型,评价多元逐步回归法优选波长的有效性。结果表明,多元散射校正技术有效降低了散射的影响,提高了相关光谱的信噪比,模型决定系数从0.598提高至0.681。基于手动挑选波长建立的模型决定系数高达0.956,预测样品集模型预测值与实测值之间的相关系数为0.823。手动挑选波长建立的SOC模型预测能力优于基于逐步回归方法建立的定量模型,利用逐步回归分析法优选波长还需要进一步研究。图7,参26。

关 键 词:近红外  多元散射校正  多元逐步回归  波长选择

Multiplicative Scatter Correction and Stepwise Regression to Build NIRS Model for Analysis of Soil Organic Carbon Content in Black Soil
SHEN Yan,ZHANG Xiao-ping,LIANG Ai-zhen,SHI Xiu-huan,FAN Ru-qin,YANG Xue-ming. Multiplicative Scatter Correction and Stepwise Regression to Build NIRS Model for Analysis of Soil Organic Carbon Content in Black Soil[J]. System Sciemces and Comprehensive Studies In Agriculture, 2010, 26(2): 174-180
Authors:SHEN Yan  ZHANG Xiao-ping  LIANG Ai-zhen  SHI Xiu-huan  FAN Ru-qin  YANG Xue-ming
Affiliation:1. Northeast Institute of Geography and Agricultural Ecology, Chinese Academy of Sciences, Changchun 130012, China; 2. Gradutute University of Chinese Academy of Sciences, Beijing 100049, China; 3. Greenhouse & Processing Crops Research Centre, Agriculture & Agri-Food Canada, Harrow, Ontario, Canada NOR 1 GO)
Abstract:Near infrared modeling based on the whole spectroscopy takes much work and the model may not be optimum because of too much tedious information. Wavelength selection through certain method could improve the model. Stepwise regression is usually used to select wavelengths with important information. Multiple scatter correction (MSC) technique can be used effectively to remove the effect of scatterings due to physical factors such as the density and humidity of samples and other factors caused by operators. A total of 136 Black Soil samples were obtained during 2004-2005 in Northeast China and corresponding infrared spectra from 3699 -12000 cm^-1 was measured using Fourier transform infrared spectrometry. Multiple scatter correction was used to preprocess the original spectra and multiple stepwise regression method was used to select wavelengths. Results showed that MSC can effectively decrease the scattering effect and thus enhance the signal to noise ratio. Compared to original spectrum, MSC improved the SOC quantitative model, with the determination coefficient increased from 0. 598 to 0.681. Determination coefficients of model based on wavelengths chosen by experience was 0.956. Model based on wavelengths selected by band was better than that based on wavelengths selected by stepwise regression analysis. Selecting wavelengths by stepwise regression analysis needs to be further studied.
Keywords:near infrared spectroscopy  multiple stepwise regression  wavelength selection
本文献已被 维普 万方数据 等数据库收录!
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号