摘要
利用NIR高光谱仪(光谱范围900~1700nm)对土壤含盐量进行了无损检测,对比分析不同含盐量土壤的水分变化情况、不同时间下土壤光谱曲线的差异。结果表明,随着土壤中含盐量的增加,土壤中水分蒸发情况受到的影响程度不同,从而使得不同含盐量土壤的反射率存在明显的规律;在此基础上,对比分析了不同预处理方法,优选出原始光谱;利用多元线性回归(multiple linear regression,MLR)、主成分回归(principal component regression,PCR)与偏最小二乘回归(partial least squares regression,PLSR)方法对900~1 700 nm范围的特征波长建立模型,对比分析不同建模效果,优选β系数提取的特征波长的PLSR模型,特征波长为936、996、1 016、1 136、1 151、1 186、1 273、1 395、1 425、1 458、1 535、1 642 nm,最优模型的预测相关系数为0.949,预测均方根误差为2.914 g/kg。因此,今后可采用不同波段对土壤含盐量进行定量分析,为今后表层土壤含盐量遥感预测奠定基础。
This article summarizes a near-infrared hyperspectral imaging technique was investigated for non-destructive determination of soil salinity,and the changes of soil moisture and soil spectral curves on different days were compared and analyzed.The results show that the evaporation of soil water is affected to different degrees with the increase of soil salinity,so that the reflectance of soil with different salinity exists an obvious rule.On this basis,different pretreatment methods were compared and analyzed to optimize the original spectrum.MLR,PCR and PLSR modeling were used to optimize the best model for feature wavelengths.Compared with different modeling effects,optimize the PLSR model of characteristic wavelength extracted by β coefficient was obtained.The optimal characteristic wavelengths are 936,996,1 016,1 136,1 151,1 186,1 273,1 395,1 425,1 458,1 535,1 642 nm,respectively.The prediction coefficient R_p is 0.949,and the RMSEP is 2.914 g/kg.Therefore,soil salinity can be quantitatively analyzed by different bands,which lays a foundation for remote sensing prediction of soil salinity in the future.
作者
吴龙国
张瑶
王松磊
李建设
WU Long-guo;ZHANG Yao;WANG Song-lei;LI Jian-she(School of Agriculture,Ningxia University,Yinchuan 750021,China;Institute of Food Testing and Research,Yinchuan 750001,China)
出处
《光电子．激光》
EI
CAS
CSCD
北大核心
2020年第4期388-394,共7页
Journal of Optoelectronics·Laser
基金
宁夏自然基金项目(2019AAC03057)
宁夏大学自然科学基金项目(ZR18013)项目资助项目。
关键词
高光谱成像
土壤盐分
诊断机理
无损检测
hyperspectral imaging
soil salinity
diagnosis mechanism
non-destruction detection
