Remote Sensing Estimation Of Sowing Date Based On Spectral Information Of Winter Wheat In The Early Growth Stage

Sowing date is an important factor affecting yield and quality of winter wheat.Timely and accurate estimation of sowing date is crucial to the production management of winter wheat.At present,based on vegetation phenology remote sensing monitoring method,using the whole growth period of remote sensing data to fit the sowing date will cause time lag.The monitoring precision caused by weak vegetation spectral signal is not high when using the early growth spectrum of winter wheat to monitor sowing date merely.In this paper,in order to realize the early monitoring of winter wheat sowing time and improve the precision of remote sensing monitoring of sowing date,multitemporal hyperspectral data of winter wheat in the early growth stage was mined,the mechanism and variation of spectral response of winter wheat at different sowing dates in the early stage of growth were discussed.The optimum phase and spectrum for estimating sowing date were also studied.The main contents and results of this thesis are as follows:(1)The mechanism and variation of canopy spectral difference of winter wheat at different sowing dates were analyzed by using the hyperspectral data measured on the ground in the early stage of winter wheat growth(from seedling emergence to jointing stage).According to the analysis,the phase of seeding date estimating can be selected after Wintering and before Regreening.(2)Based on the principle of maximization of spectral difference in canopy of winter wheat at different sowing dates,the optimal timing for estimating sowing date was obtained.Based on the principle of the most relevant to the sowing date,the optimal spectrum for sowing period estimating was selected.The results showed that the sowing date and spectral index of November 8th 2016 were the best,and early November was the best time phase for estimating sowing date of winter wheat in Beijing.Based on the derivative spectra of the red and blue edge spectral variation regions,the spectral feature edge parameters(SDr-SDb)/(SDr+SDb)are optimized in the estimation of sowing time.(3)Based on the ground measured hyperspectral data and UAV multi-spectral image data,the optimal seeding date was estimated and the optimal estimation model of seeding time was constructed.The results showed that,HSV colorimetric transformation method could extract winter wheat accurately.PVI was the optimal index for estimating the sowing date of winter wheat in 14 spectral indices.Based on grey relational analysis and the Akaike's information criterion,the optimal winter wheat sowing date model was the PLS1 partial least squares regression model constructed by RGRI? Datt4 ? Datt6 and SRre independent variables.Compared with the multivariate regression model,the estimation accuracy of the stochastic forest algorithm was higher.It is a new remote sensing method for sowing date estimation of winter wheat and has high application value.