文章摘要
倪 斌.基于CASI & SASI航空高光谱的雄安新区西南部农田土壤 重金属镍含量反演研究[J].地质与勘探,2022,58(6):1307-1320
基于CASI & SASI航空高光谱的雄安新区西南部农田土壤 重金属镍含量反演研究
Retrieval of heavy metal nickel content in farmland soil in the southwest of Xiong’an New District based on aerial hyperspectral CASI & SASI data
投稿时间:2021-12-08  修订日期:2022-03-26
DOI:
中文关键词: CASI&SASI 高光谱 多元逐步回归 偏最小二乘回归 神经网络 重金属反演 模型评价 雄安新区
英文关键词: CASI & SASI, hyperspectral data, SMLR, PLSR, NN, heavy metal inversion, model evaluation, Xiong’an New District
基金项目:国家重点研发计划(编号:2016YFC0600210)、国家自然科学基金(编号:41272366;41401515)和中国冶金地质总局科技创新项目(编号:CMGB202001)联合资助
作者单位E-mail
倪 斌 中国冶金地质总局矿产资源研究院北京 hzhaoq@126.com 
摘要点击次数: 210
全文下载次数: 226
中文摘要:
      农田土壤中重金属元素富集会严重制约农作物的生长,且对人类健康造成潜在威胁。高光谱遥感数据具有极高的光谱分辨率,因而可在土壤重金属污染元素信息的定量研究中发挥重要作用。本文以雄安新区西南部及其周边农田土壤作为研究对象,在实验室测定土壤重金属元素Ni的含量,并与土壤可见-近红外高光谱数据建立土壤重金属Ni含量的定量估测模型,进一步基于CASI&SASI航空高光谱数据快速反演研究区农田土壤重金属Ni的含量,获取其分布特征。本文研究并建立了研究区土壤重金属元素基于不同光谱变换形式的多元逐步回归、偏最小二乘回归和BP神经网络统计估算模型,通过模型验证与对比,探索研究区土壤重金属Ni元素含量的最优反演模型。研究结果表明: (1)基于各光谱变换的BP神经网络模型的建模和预测精度整体上大于偏最小二乘法和多元逐步回归法模型,模型拟合精度高,预测能力较好;(2)综合来看,一阶微分处理能普遍改善模型预测效果,其中BP神经网络模型的一阶微分变换结果最佳,对于Ni元素建模精度R2高达97.1%,验证集精度R2高达98%以上;(3)选用精度最好的BP神经网络模型,通过CASI&SASI高光谱数据对研究区重金属Ni含量进行反演,反演结果与实测Ni含量数据一致性很好。
英文摘要:
      The enrichment of heavy metal elements in farmland soil will seriously restrict the growth of crops and pose a potential threat to human health. Hyperspectral remote sensing can well obtain the quantitative information of soil heavy metal pollution elements due to its high spectral resolution. Taking the farmland soil in the southwest of Xiong'an New District and its surrounding as an example, this work measured the contents of heavy metal elements Ni in the soil in the laboratory, and established quantitative estimation models for Ni content in combination with soil visible-near infrared hyperspectral data. Based on CASI & SASI airborne hyperspectral data, the content distribution of heavy metal Ni in farmland soil in the study area was quickly obtained. This study constructed statistical estimation models for determining soil heavy metal elements, which employed different spectral transform methods of stepwise multiple linear regression (SMLR), partial least squares regression (PLSR) and back propagation neural network (BPNN). Verification and comparison of the models were conducted to explore the optimal inversion model of soil heavy metal Ni content in the study area. Results show that: (1) The modeling and prediction accuracy of the BPNN model based on spectral transformation are better than those of the PLSR and SMLR models, featured by high fitting accuracy and good prediction ability. (2) On the whole, the first derivative processing can generally improve the prediction effect of the model, among which the first derivative transformation result of BPNN model is the best. For Ni element, the modeling accuracy R2 is as high as 97.1% and the verification set accuracy R2 is greater than 98%. (3) The BPNN model with the best accuracy was selected to inverse the Ni content in the study area by using CASI & SASI hyperspectral data. The inversion results are in good agreement with the measured Ni content data.
查看全文   查看/发表评论  下载PDF阅读器
关闭