Study On Hyperspectral Monitoring Of Agronomic Parameters Of Millet Based On Wavelet Transform And Least Squares

The existing models for estimating the agronomic parameters of foxtail millet are easily affected by region,growth period and environmental factors,and the spectral curve is noisy and the spectral information cannot be used effectively,therefore,in order to improve the prediction ability and robustness of millet estimation model,it is necessary to introduce new methods to use as much spectral information as possible.In recent years,continuous wavelet transform and partial least squares have been widely used by scholars,however,there are few researches on combining the advantages of the two methods and applying them to the establishment of estimating models for agronomic parameters of millet.In this study,millet(Jingu 21,Jingu28)was taken as the research object.The spectral reflectance and main agronomic parameters were measured at different growth stages.Then the orthogonal coif3 wavelet,db5 wavelet,Haar Wavelet,sym8 wavelet,bior1.5 wavelet and rbior3.1 wavelet are selected to transform the canopy spectrum and extract the wavelet energy coefficient,and carry on the correlation analysis with each agronomy parameter,selects the performance best wavelet base to carry on the different scale decomposition,determines the optimal decomposition scale,then,the wavelet energy coefficient estimation model of foxtail millet agronomic parameters was established by using partial least squares and validated.The results show that:1.The change trend of canopy spectra of foxtail millet under different organic fertilizer treatments was consistent.With the advance of growth period and the change of organic fertilizer application amount,the spectral reflectance shows great difference.With the increase of organic fertilizer application,the peak value of "green peak" decreased,and the near-infrared band showed an increasing trend.The agronomic parameters of foxtail millet were significantly changed after different organic fertilizer treatments.The plant height and chlorophyll content increased first and then decreased with the advancing of growth period,and the nitrogen content of foxtail millet increased gradually,the change of parameters in different growth stages was consistent,and increased with the increase of organic fertilizer application2.After processing by different wavelet bases,the correlation between Wavelet Energy Coefficients and agricultural parameters is analyzed,and the wavelet energy coefficients are decomposed at different scales,the correlation of Agronomic parameters of foxtail millet was improved by different wavelet bases at different decomposition scales.The correlation between bior1.5 wavelet scale 8 and plant height was the best,The db5 wavelet(scale=8),rbio3.1 wavelet(scale=2)and coif3 wavelet(scale=64),sym8 wavelet(scale=8)and Haar wavelet(scale=8)had the best correlation with plant nitrogen.3.Compared with the original spectrum,the model established by wavelet transform is better than the original spectrum.When the decomposition scale of bior1.5 and db5 wavelets is 8,the model constructed by wavelet coefficients is the best.For coif3 wavelet,the model constructed by wavelet coefficients is the best when the decomposition scale is 64.The results show that the model has better prediction ability and robustness after wavelet transform,which makes the model have more application value.