基于CWT-sCARS的土壤铜含量高光谱反演  

Hyperspectral Inversion of Soil Cu Content Based on CWT-sCARS

作　　者：张世文[1] 李唯佳 李恩伟 朱曾红 孔晨晨 ZHANG Shi-wen;LI Wei-jia;LI En-wei;ZHU Zeng-hong;KONG Chen-chen(School of Earth and Environment,Anhui University of Science and Technology,Huainan,232001,Anhui;School of Spatial Informatics and Geomatics Engineering,Anhui University of Science and Technology,Huainan,232001,Anhui)

机构地区：[1]安徽理工大学地球与环境学院,安徽淮南232001 [2]安徽理工大学空间信息与测绘工程学院,安徽淮南232001

出　　处：《蚌埠学院学报》2024年第2期17-23,共7页Journal of Bengbu University

基　　金：塔里木河流域土地开发与农业资源调查(2021xjkk0200);淮北矿业集团科技研发项目(2022-103)。

摘　　要：光谱变量的有效程度与土壤铜含量的反演精度密切相关。基于原始反射率以及不同分解尺度下的小波系数,本研究采用连续小波变换(CWT)算法、稳定性竞争自适应重加权采样(sCARS)算法和随机森林(RF)算法对土壤铜含量进行了反演与验证。研究结果表明:连续小波变换可以有效提高光谱特征与土壤铜含量之间的相关性,不同分解尺度对应的最大相关系数中,最大值位于Scale 8分解尺度下1343 nm处,相关系数为0.60;使用sCARS算法可以显著减少特征变量的数量,结合CWT变换和sCARS算法可以显著减轻数据冗余,提高土壤Cu含量的反演精度。该研究可为利用高光谱遥感技术,快速、高精度反演土壤Cu含量提供重要参考。The practical degree of spectral variables is closely related to the inversion accuracy of soil copper content.Based on the original reflectance and the wavelet coefficients at different decomposition scales,the continuous wavelet transform(CWT)algorithm,the stability-competitive adaptive reweighted sampling(sCARS)algorithm,and the random forest(RF)algorithm were used in this study to invert and validate the soil copper content.The results showed that the continuous wavelet transform can effectively improve the correlation between spectral features and soil copper content.Among the maximum correlation coefficients corresponding to different decomposition scales,the maximum value is located at 1343 nm under the Scale 8 decomposition scale,with a correlation coefficient 0.60.The use of the sCARS algorithm significantly reduces the number of feature variables.Combining the CWT transform and the sCARS algorithm can significantly reduce the data redundancy and improve the inversion accuracy of the soil Cu content.It can provide an essential reference in this study for the rapid and high-precision inversion of soil Cu content using hyperspectral remote sensing technology.

关 键 词：高光谱反演 连续小波变换 稳定性竞争自适应重加权采样 

分 类 号：S127[农业科学—农业基础科学]