The practical degree of spectral variables is closely related to the inversion accuracy of soil cop-per 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 un-der 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 algo-rithm can significantly reduce the data redundancy and improve the inversion accuracy of the soil Cu con-tent .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
Key words
hyperspectral inversion/continuous wavelet transform/stability of competitive adaptive reweighted sampling
维普
万方数据