Estimation Of Soil Cu Content By Hyperspectral Fractional Differential Transformation

The problem of soil heavy metal pollution is becoming seriously,which has become a worldwide pollution problem.At present,it is a new development trend to monitor soil heavy metal pollution by quantitative remote sensing.The thesis selected farmland soil locate at a certain area of Yueyang as research object,and the spectral reflectance data of the sample soil were processed by fractional differential treatment of 0?2 order based on the measured hyperspectral data and the soil heavy metal detection data,and combine with related coefficient indicators to explore effects of fractional differential on high spectral data of soi1 heavy metals within different sampling intervals and different mathematical transformations,based on this,under different sampling intervals and different mathematical transformations,based on fractional order differential,we establish a high spectral estimation model of heavy metals.The main research result of this thesis as follows:(1)For different sampling interval resampling,after 0?2 order fractional differential(interval 0.2 order,total 11 orders),the research shows,as the differential order increases,the number of bands that pass the extremely significant test increases has increased significantly,the trend of the number of bands increasing first and then decreasing;the maximum of absolute value of correlation coefficient |P| max increases first and then decreases,and the absolute value of correlation coefficient in15nm reaches the maximum at 535nm of 1.6 order differential,under fractional differential,in all sampling intervals,which is the maximum value of |P| max;the effect of increasing correlation at 15 nm resampling interval is the best among the seven sampling intervals.(2)For different mathematical transformation,after 0?2 order fractional differential(interval 0.2 order,total 11 orders),the study founds,compared to the four transform spectra,the correlation between the original spectrum and heavy metal Cu element is the strongest,in original spectrum,|P| max reaches the maximum at 535nm of 1.6 order differential.(3)In this study,the partial least squares regression method(PLSR)was used to establish the estimation model of soil heavy metal Cu content,the results show that the optimal model for this study is a PLSR model which be with a 15 nm sampling interval and a 1.6-order fractional differential under the reciprocal transformation,the R2Y of the model is 0.678,RMSEY is 1.702 and RPD is 1.564.In this study,fractional derivative was introduced into the pretreatment of hyperspectral data of heavy metals in soil.It can be used to retrieve the content of soil heavy metals from hyperspectral data,and to monitor large-scale heavy metals pollution using space-borne hyperspectral data.It can also provide a useful reference for the band setting scheme of remote sensing satellite sensors for resources.