水分对近红外光谱检测紫色土有机质含量影响的研究  

A Research of the Effects of Soil Moisture on NIR Spectroscopic Detection of Organic Matter Content in Purple Soil

作　　者：曾静[1] 李光林[1] 周胜灵[1] 

机构地区：[1]西南大学工程技术学院,重庆400715

出　　处：《西南大学学报（自然科学版）》2015年第10期167-173,共7页Journal of Southwest University(Natural Science Edition)

基　　金：重庆市科委项目(cstc2014yykfA);中央高校基本科研业务费专项(2362014xk13)

摘　　要：近红外光谱（NIR）技术在线检测土壤养分含量是精细农业研究热点之一.土壤水分在近红外波段具有较高的吸收系数,处于波动状态的土壤水分,直接影响NIR技术检测土壤有机质含量的预测精度.四川、重庆地区紫色土中有机质含量较低,比土壤水分最大有一个数量级的差异.利用BRUKER公司的MPA傅里叶近红外谱仪对不同含水率的紫色土壤进行了近红外光谱扫描,分析水分对近红外光谱检测紫色土壤有机质含量模型精度影响,提出了一种消除土壤水分对紫色土壤有机质含量预测影响的方法.近红外光谱扫描显示,在同一有机质含量水平下,含水率增加,光谱吸光度呈非线性上升趋势,直接利用光谱数据建立的偏最小二乘法（PLS）模型,在较高水分条件下,模型精度低;在遗传算法选择的波数区域,用二次多项式拟合紫色土壤含水率指数对近红外扫描光谱吸光度影响,其相关系数在0.95-0.97之间;将一定含水率下的近红外光谱吸光度折算为干土条件下吸光度,再利用PLS法建立近红外光谱有机质含量预测模型,与原有模型相比预测精度得到显著提高.实验结果表明,本方法可以有效消除紫色土壤水分对近红外光谱检测其有机质含量预测影响,为不同水分条件下近红外光谱在线检测紫色土壤有机质含量提供一定的理论和技术支持。On-line detection of soil nutrient content with NIR reflectance spectroscopy is one of the research hotspots of precision agriculture.Not only does soil moisture have a higher absorption coefficient in the near infrared wavelengths,its content also fluctuates considerably.Consequently,these factors have a direct impact on the prediction accuracy of NIR reflectance spectroscopic detection of soil organic matter content.In Sichuan and Chongqing,the purple soil has a low level of soil organic matter content,and soil moisture and organic matter contents have a magnitude difference.In this study,the Fourier infrared spectrometer of MPA BRUKER Company was used to conduct near infrared spectroscopy scanning of different water moisture contents of purple soil,and the effect of moisture on the accuracy of the determination was analyzed,based on the results of which a method to eliminate the effect of soil moisture on forecasting soil organic matter content of purple soil was presented.NIR spectrometer scanning showed that under the same level of organic matter content,the spectral absorbance increased nonlinearly with increasing water content.Consequently least squares models established by directly using spectroscopic data had low model accuracy in high moisture conditions,which cannot satisfy the use.In the area of genetic algorithm of wave number,using quadratic polynomial to fit the effects of the purple soil moisture on the spectral absorbance of NIR scanning,and the correlation coefficients varied between 0.95 and 0.97.The measurements of absorbance of NIR spectroscopy under a certain moisture level were converted into those under dry soil conditions,and the PLS method was used to establish an NIR spectral forecast model.Compared with the original model,the newly established model greatly improved the forecasting accuracy.The above results showed that this approach may effectively eliminate the influence of purple soil moisture on NIR spectroscopy detection of the organic matter content,and provide certain theoretic

关 键 词：水分影响 近红外光谱技术 光谱吸光度处理 紫色土有机质 精细农业 

分 类 号：S158.2[农业科学—土壤学]