变换光谱数据对土壤氮素PLSR模型的影响研究

发布时间：2018-11-06 19:58

【摘要】：光谱数据变换对消除背景、噪音影响以及提取光谱特征有重要的作用,是光谱数据分析过程中的必要步骤。为了研究光谱变换处理对土壤氮素PLSR模型的影响精度,并选择最佳光谱变换处理方法,本文对原始光谱数据进行了15种典型光谱变换,通过比较不同变换光谱与土壤氮素的相关性,实现土壤氮素的PLSR精确诊断,并综合评定最佳光谱数据变换方法。结果表明,涉及微分处理后的光谱变换,尤其是先进行开方(T8、T11)、对数(T6、T12)等变换后再进行微分处理,可提高其与土壤氮素的相关性。在引入较少因子变量个数的条件下,该方法使因变量解释量达到了98%。综合考虑模型的校正、验证效果及模型复杂度(模型最佳因子变量个数),可得出光谱平方根的一阶微分变换处理(T8)为最佳的土壤光谱变换算法。该条件下的土壤氮素的校正模型表现为R2=0.985、RMSEC=0.000132、Fn=6,验证模型的表现为R2=0.9853、RMSEV=0.000162,结果表明基于T8的光谱数据变换可实现本试验条件下土壤氮素的光谱估算。另外,可以考虑将原始光谱的一阶微分(T9)、对数和对数倒数的一阶微分(T6、T7)以及平方根和对数的二阶微分(T11、T12)作为光谱数据变换方法。本文研究结果可为土壤氮素估算和光谱数据预处理提供技术参考。
[Abstract]:Spectral data transformation plays an important role in eliminating background noise and extracting spectral features. It is a necessary step in the process of spectral data analysis. In order to study the effect of spectral transformation treatment on soil nitrogen PLSR model and select the best spectral transformation processing method, 15 typical spectral transformations were carried out for the original spectral data in this paper. By comparing the correlation between different transformation spectra and soil nitrogen, the accurate diagnosis of soil nitrogen was realized by PLSR, and the optimal spectral data transformation method was evaluated synthetically. The results showed that the correlation between the spectral transformation after differential treatment and soil nitrogen could be improved, especially when the formula (T8 T11) and logarithm (T6 T12) were used before differential treatment. Under the condition that the number of factor variables is less, the interpretation quantity of dependent variables reaches 98. Considering the correction of the model, the verification effect and the complexity of the model (the number of the best factor variables of the model), it is concluded that the first order differential transform (T8) of the spectral square root is the best algorithm for soil spectral transformation. The calibration model of soil nitrogen under this condition was shown as R2N 0.985N RMSEC 0.000132FnN 6, and the validation model was R2N 0.9853 RMSEV 0.000162. The results showed that the spectral data transformation based on T8 could realize the spectral estimation of soil nitrogen under this experimental condition. In addition, the first order differential (T9) of the original spectrum, the first order differential of logarithmic and logarithmic inverse (T6T7) and the second order differential of square root and logarithm (T11T12) can be considered as spectral data transformation methods. The results can provide technical reference for soil nitrogen estimation and spectral data preprocessing.
【作者单位】：山西农业大学旱作农业工程研究所;
【基金】：国家自然科学基金项目(31371572;31201168) 山西省科学技术发展计划项目(201603D221037-3) 山西省归国人员重点资助项目(2014-重点4) 山西省科技攻关项目(20110311038) 山西省青年基金项目(2012021023-5)
【分类号】：S151.93

【相似文献】