论文部分内容阅读
通过对新疆维吾尔自治区温宿县、和田县、拜城县191个土样的原始反射率进行连续统去除及连续统去除的一阶微分处理,分析了盐分的高光谱吸收特征及敏感波段,并建立了盐分含量的多种高光谱定量反演模型。结果表明,在400~2 400 nm波段,土壤反射率与盐分含量之间无明显规律。640~700 nm波段的连续统去除数据与含盐量呈极显著负相关,而710~780 nm波段呈极显著正相关。原始反射率经连续统去除处理后,可明显提高反演模型的预测性能。基于盐分光谱指数和吸收特征参数构建的反演模型的稳定性及预测能力不如连续统去除、连续统去除一阶微分的400~2 400 nm或敏感波段的偏最小二乘回归(PLSR)模型。所有模型中,仅有以400~2 400 nm和640~700 nm连续统去除数据所建模型的相对分析误差(RPD)达2.5以上,分别为2.62和2.52,且二者其余各项评价指标差异不大。以640~700 nm波段连续统去除数据构建的PLSR模型对南疆水稻土盐分含量具有很好的反演效果。
Based on the first order differential processing of the original reflectance of 191 soil samples in Wensu County, Hetian County and Baicheng County, Xinjiang Uygur Autonomous Region, the hyperspectral absorption characteristics of salt and the sensitive bands were analyzed. A series of hyperspectral quantitative inversion models of salt content were established. The results showed that there was no obvious regularity between soil reflectance and salt content in the 400-2 400 nm band. The data from 640 to 700 nm were significantly and negatively correlated with salinity, while the bands from 710 nm to 780 nm showed a significant positive correlation. After the original reflectivity is removed by continuous removal, the prediction performance of the inversion model can be obviously improved. The stability and the prediction ability of the inversion model based on the salt spectral index and the absorption characteristic parameters are not as good as that of the continuum removal and the PLSR model of 400 ~ 2 400 nm or sensitive bands of first-order differential. In all the models, only the relative analysis error (RPD) of the model built by continuous data removal at 400-2 400 nm and 640-700 nm was more than 2.5, which were 2.62 and 2.52, respectively, and the rest of the evaluation indexes were different Not big. The PLSR model constructed with data from 640 to 700 nm is a good inversion of the salt content in the southern paddy soil.