基于sCARS的淮北平原土壤有机质含量高光谱建模  

Based on sCARS Hyperspectral Modeling of Soil Organic Matter Content in Huaibei Plain

作　　者：高迎凤 赵明松[1,2,3] 于芝琳 赵治东 王涛 GAO Ying-feng;ZHAO Ming-song;YU Zhi-lin;ZHAO Zhi-dong;WANG Tao(School of Geomatics,Anhui University of Science and Technology,Huainan 232001,China;Key Laboratory of Aviation-Aerospace-Ground Cooperative Monitoring and EarlyWarning of Coal Mining-Induced Disasters of Anhui Higher Education Institutes,Huainan 232001,China;Coal Industry Engineering Research Center of Collaborative Monitoring of Mining Area’s Environment and Disasters,Huainan 232001,China)

机构地区：[1]安徽理工大学空间信息与测绘工程学院,安徽淮南232001 [2]矿山采动灾害空天地协同监测与预警安徽省教育厅重点实验室,安徽淮南232001 [3]矿区环境与灾害协同监测煤炭行业工程研究中心,安徽淮南232001

出　　处：《安徽师范大学学报（自然科学版）》2024年第3期255-263,共9页Journal of Anhui Normal University(Natural Science)

基　　金：安徽省自然科学基金项目(2208085MD88);安徽理工大学人才引进项目(ZY020).

摘　　要：为确定淮北平原砂姜黑土土壤有机质(SOM)最佳反演模型,探寻最佳特征波长筛选方法,提高模型预测精度。利用原始光谱进行倒数对数(Log(1/R))、标准正态变量变换(SNV)、去包络线(CR)、一阶微分(FDR)处理,采用稳定竞争性自适应重加权算法(sCARS)筛选特征变量,对比分析竞争性自适应重加权算法(CARS)、相关系数法(|r|≥0.47)和显著性水平法(p≤0.01)所得结果,建立SOM含量的偏最小二乘(PLSR)模型,并对比精度差异。结果表明:(1)全波段范围内,SOM含量与原始光谱呈极显著负相关,与Log(1/R)光谱呈极显著正相关,与SNV光谱相关性明显增强。CR和FDR光谱与SOM含量呈不同程度的正负相关性。(2)对比全波段,CARS和sCARS算法能够有效去除光谱冗余信息,筛选得到特征波段数目仅占全波段的1%~5%。筛选后模型精度更高,相对分析误差(RPD)均大于1.8。(3)相比于CARS算法,sCARS算法具备更好的稳定性和精确性。筛选到的特征波段主要分布在800~850、1850~1900、2050~2500 nm区域。(4)Log(1/R)-sCARS模型精度最佳,建模集和预测集的决定系数(R2)分别提升了0.08和0.28,RPD值为3.05,对SOM含量预测极好。In order to determine the best inversion model of soil organic matter(SOM)of sand and Shajiang black soil in the Huaibei Plain,the best characteristic wavelength screening method was explored to improve the prediction accuracy of the model.The original spectrum was used to perform inverse-log reflectance(Log(1/R)),standard normal variable(SNV),continuum removal(CR),and first-order derivative reflectance(FDR)processing,and the stability competitive adaptive reweighted sampling(sCARS)was used to screen the charac-teristic variables,and the results obtained by the competitive adaptive reweighted sampling(CARS),correlation coefficient method(|r|≥0.47)and significance level method(p≤0.01)were compared and analyzed,and a partial least squares regression(PLSR)model of SOM content was established.And compare the accuracy differences.The results show that:(1)In the whole band,the SOM content is negatively correlated with the original spec-trum,positively correlated with the Log(1/R)spectrum,and significantly enhanced with the SNV spectrum.CR and FDR spectra showed different degrees of positive and negative correlation with SOM content.(2)Compared with the full band,the CARS and sCARS algorithms can effectively remove the spectral redundancy informa-tion,and the number of characteristic bands screened accounts for only 1%~5%of the full band.After screening,the accuracy of the model was higher,and the relative percent deviation(RPD)was greater than 1.8.(3)Com-pared with the CARS algorithm,the sCARS algorithm has better stability and accuracy.The screened characteris-tic bands are mainly distributed in the 800~850,1850~1900,2050~2500nm regions.(4)The Log(1/R)-sCARS model has the best accuracy,the determination coefficients(R2)of the modeling set and the prediction set are in-creased by 0.08 and 0.28,respectively,and the RPD value is 3.05,which is excellent for the prediction of SOM content.

关 键 词：土壤有机质 砂姜黑土 光谱变换 sCARS筛选 偏最小二乘模型 

分 类 号：TP79[自动化与计算机技术—检测技术与自动化装置] S151.9[自动化与计算机技术—控制科学与工程]