| 贾学勤,冯美臣,杨武德,王超,孙慧,武改红,张松.基于SPA-MLR方法的土壤含水量光谱预测模型研究[J].干旱地区农业研究,2018,36(3):266~269 |
| 基于SPA-MLR方法的土壤含水量光谱预测模型研究 |
| Study on the spectral prediction model of soil moisture content based on SPA-MLR method |
| |
| DOI：10.7606/j.issn.1000-7601.2018.03.39 |
| 中文关键词: 光谱变换 土壤含水量 连续投影算法 多元线性回归 |
| 英文关键词:spectral transform soil moisture content successive projection algorithm multiple linear regression |
| 基金项目:山西省归国人员重点资助项目（2014-重点4）；山西省科学技术发展计划项目（201603D221037-3）；山西省科技攻关项目（20110311038） |
|
| 摘要点击次数: 1352 |
| 全文下载次数: 596 |
| 中文摘要: |
| 以人工调配的不同含水量土壤的高光谱数据为基础，运用11种常规的变换方法对原始光谱反射率进行变换，使用连续投影算法（SPA）提取特征波段，然后建立多元线性回归（MLR）模型，并对不同模型进行评价比较，旨在选择监测土壤含水量的最佳高光谱模型，实现土壤含水量高光谱监测。结果表明，随着土壤含水量的增加光谱反射率先升高后降低；使用SPA提取的特征波段为3～5个，且不同变换处理后提取的特征波段存在差异。利用特征波段建立MLR回归模型，表明原始光谱经一定数学变换处理可以提高土壤含水量高光谱监测精度，其中对数的一阶微分变换处理（T8）后建立的SPA-MLR模型监测精度最高，其校正模型表现为R2=0.957，RMSE=2.16，RPD=4.74，验证模型表现为R2=0.903，RMSE=3.41，RPD=2.95。故基于反射率对数一阶微分变换处理所建立的SPA-MLR模型可以更好地实现土壤含水量的高光谱监测。 |
| 英文摘要: |
| Based on hyperspectral data of artificially deployed soil with different moisture content, 11 conventional transformation methods were used to transform the original spectral reflectance, successive projection algorithm (SPA) was used to extract the sensitive wavelengths, and then the multiple linear regression (MLR) model was established. different models were evaluated and compared in order to select the best hyperspectral model for monitoring soil moisture and achieve hyperspectral monitoring of soil moisture content. The results showed that the spectral reflectance increased first and then decreased with the increase of soil moisture content; and the characteristic bands of SPA extraction ranged from 3 to 5, and there were differences in the characteristic bands extracted by different spectral transformations. The establishment of the MLR regression model using the characteristic wavebands shows that the original spectrum can improve the hyperspectral monitoring accuracy of soil moisture after a certain mathematical transformation. the SPA[CD*2]MLR model based on spectral reflectivity of the first-order differential logarithmic transform (T8) were the best, the calibration model showed that R2=0.957，RMSE=2.16，RPD=4.74 and validation model showed that R2=0.903，RMSE=3.41，RPD=2.95. Therefore, the SPA-MLR model based on the logarithmic first differential transformation of reflectance can realize the hyperspectral monitor of soil moisture content, and the study would provide technical support for rapid monitoring of the soil moisture content. |
| 查看全文 查看/发表评论 下载PDF阅读器 |