干旱区生态环境知识资源中心(Arid): 基于分数阶微分的博斯腾湖湖滨绿洲土壤电导率高光谱估算

Arid

	基于分数阶微分的博斯腾湖湖滨绿洲土壤电导率高光谱估算
其他题名	Fractional differential based hyperspectral estimation model of soil electrical conductivity of Bosten Lake oasis
	赵慧; 李新国; 靳万贵; 麦麦提吐尔逊·艾则孜; 牛芳鹏
来源期刊	甘肃农业大学学报
ISSN	1003-4315
出版年	2021
卷号	56 期号:1 页码:1003-4315
中文摘要	【目的】探讨分数阶微分在高光谱估算湖滨带绿洲土壤电导率的可行性,实现土壤盐分的高精度监测. 【方法】以博斯腾湖湖滨绿洲为研究区,引入分数阶微分对土壤高光谱反射率进行0~2阶微分处理,利用相关性检验和连续投影算法优选特征波段,构建土壤电导率的全波段和特征波段,以及利用连续投影算法优选的全波段和特征波段的4种偏最小二乘回归模型,依据决定系数、均方根误差优选土壤电导率的最佳估算模型.【结果】通过P=0.01显著性检验的特征波段,随着阶数的增加,土壤电导率的特征波段数从725增加至776再减少至498;相关系数值随着阶数的增加从0.48至0.68再减少至0.59.0~0.6阶的相关系数绝对值最大值对应的特征波段分布在581~744nm,0.8阶~2阶相关系数绝对值最大值对应的波段分布在近红外波段.利用连续投影算法筛选土壤电导率的全波段,其1.4阶微分的波段数量最多,达到6个;利用连续投影算法筛选土壤电导率的特征波段,其2阶微分的波段数量最多,达到3个.连续投影算筛选全波段的偏最小二乘回归模型为土壤电导率的最优模型,分数阶为1.6阶,波段分布在近红外波段,R_C~2 =0.64,RMSE_C=3.06,R_V~2 =0.83,RMSE_V =2.86.【结论】基于分数阶微分的高光谱估算模型具有较好的稳定性,模型估算的预测能力较好,可以为湖滨带绿洲土壤高光谱估算土壤电导率含量提供方法支持.
英文摘要	【Objective】In order to achieve high-precision monitoring of soil salinity,the feasibility of estimating soil electrical conductivity in the lakeside oasis through fractional differential was studied. 【Method】The lakeside oasis on the west coast of Bosten Lake was selected as the research area,the fractional differential was introduced to differentiate the spectral reflectance from 0to 2,and the characteristic band was selected by a statistical correlation test.The continuous projection algorithm was used to screen the full band and the characteristic band.Four partial least squares regression models of full-band,characteristic band,full-band and feature bands based on continuous projection algorithm were selected,and the optimal inversion model was optimized according to the decision coefficient,root means square error and relative analysis error.【Result】With the increase of fractional differential,the number of passing P=0.01 significance test characteristic bands of soil electrical conductivity increased from 725to 776and then to 498,and the correlation value increased from 0.48to 0.68and then to 0.59.The characteristic bands corresponding to the absolute value of correlation coefficients of order 0~0.6were distributed at 581~744nm, and the bands corresponding to the maximum absolute value of correlation coefficients of order 0.8~2 were distributed at near-infrared bands.Using the continuous projection algorithm to screen the full-band, the number of bands with 1.4order differential was the largest(6);using the continuous projection algorithm to filter the feature band,the number of bands with the second-order differential was the largest(3). The partial least-squares regression model screened by continuous projection calculation was the best for estimating soil electrical conductivity,with a fractional order of 1.6.The bands distributed in near-infrared bands(R2 C=0.64,RMSE_C=3.06,R_V~2 =0.83,RMSE_V=2.86).【Conclusion】The hyperspectral estimation model based on fractional differential indicates good stability and prediction ability,which can provide methodology support for hyperspectral estimation of soil electrical conductivity in lakeside oasis.
中文关键词	土壤电导率 ; 分数阶微分 ; 连续投影算法 ; 全波段 ; 近红外光谱反演
英文关键词	soil conductivity fractional differential continuous projection algorithm full band spectrum near-infrared spectral inversion
类型	Article
语种	中文
国家	中国
收录类别	CSCD
WOS类目	Remote Sensing
CSCD记录号	CSCD:6939786
来源机构	新疆师范大学
资源类型	期刊论文
条目标识符	http://119.78.100.177/qdio/handle/2XILL650/354240
作者单位	赵慧, 新疆师范大学地理科学与旅游学院, 新疆干旱区湖泊环境与资源实验室, 乌鲁木齐, 新疆 830054, 中国.; 李新国, 新疆师范大学地理科学与旅游学院, 新疆干旱区湖泊环境与资源实验室, 乌鲁木齐, 新疆 830054, 中国.; 靳万贵, 新疆师范大学地理科学与旅游学院, 新疆干旱区湖泊环境与资源实验室, 乌鲁木齐, 新疆 830054, 中国.; 麦麦提吐尔逊·艾则孜, 新疆师范大学地理科学与旅游学院, 新疆干旱区湖泊环境与资源实验室, 乌鲁木齐, 新疆 830054, 中国.; 牛芳鹏, 新疆师范大学地理科学与旅游学院, 新疆干旱区湖泊环境与资源实验室, 乌鲁木齐, 新疆 830054, 中国.
推荐引用方式 GB/T 7714	赵慧,李新国,靳万贵,等. 基于分数阶微分的博斯腾湖湖滨绿洲土壤电导率高光谱估算[J]. 新疆师范大学,2021,56(1):1003-4315.
APA	赵慧,李新国,靳万贵,麦麦提吐尔逊·艾则孜,&牛芳鹏.(2021).基于分数阶微分的博斯腾湖湖滨绿洲土壤电导率高光谱估算.甘肃农业大学学报,56(1),1003-4315.
MLA	赵慧,et al."基于分数阶微分的博斯腾湖湖滨绿洲土壤电导率高光谱估算".甘肃农业大学学报 56.1(2021):1003-4315.