The Study On Estimate Models For Cotton Yield Component And Quality By Hyperspectral Remote Sensing

[Research object] In this essay, based on different treatments of cotton in XinJiang, using hyperspectral remote sensing, cotton canopy have been multi-time monitord. Based on the spectral analysis techniques and a variety of analysis algorithms, analysis cotton yield component, leaf Biochemistry, fiber quality indicators and high spectral data correlation ,choose the best characteristic parameters of hyperspectral and vegetation index, a yield component and the best indicators of fiber quantitative model of cotton will be established.[Research methods] This study investigated the yield and quality of cotton a crucial period canopy reflectance spectra of two analytical methods used. first, analysis the correlation coefficients between the spectral reflectance ,vegetation index and yield component or fibre quality by multivariate statistical analysis and stepwise regression analysis. Second, based on canopy spectral variable analysis techniques, choose the better spectral parameters and biochemical parameters to established a fitting cotton production and quality monitoring mathematical regression model.[Research results]1, Correlation between cotton yield components and canopy reflectance reached significant level. Normalization of vegetation index [920,980nm] was proved to be able to predict the total boll number per area. Using spectral vegetation index [820,1650nm] can be estimated cotton boll weight in different stages. Normalization of vegetation index [820,1650nm] was proved to be able to predict the seed index in boll-opening stages. Through prediction model test,there is a possible to use P_Depth1108 to estimated lint index at later peaking bolling stages.2, Through directly spectral monitor fiber,the result showed that boll-opening band sensitive range is richer than later peaking bolling stage. We can choose Depth1931 spectral characteristics of the common parameters to predict fiber quality indicators in directatly; Boll-opening stage, we can use common parameters of the best selected P_Depth1112 to predict the major cotton fiber quality indicators.3, Through hyperspectral data to estimate leaf nitrogen content/ the quality indicators of cotton.The result showed that there was a well relationship about leaf nitrogen content and fiber length, strength, Micronaire Value. The sensitive band range of leaf nitrogen content is 730-940nm and 1970-2500nm. By analyzing the leaf nitrogen content and high spectral characteristic parameters, it can be used (730-940nm) spectral reflectance countdown to predict full bloom leaf nitrogen content.Based on above relationship,rhese results suggested that it was feasible to monitor and predict fibre length and intensity by using remote sensing technique.4,At full boll stages, NRI and VARI_green can be used to estimate soluble sugar content of cotton(decision coefficient = 0.5696 and 0.6295). We can useσand Depth962 to estimated the cellulose content (decision coefficient=0.6888 and 0.5766). Established the regression equation of fiber soluble sugar content (x) and fiber length (y) at early bolling stage,y=-0.1172x2+9.1655x-148.71(R~2=0.6532); the regression equation of cellulose content (x) and fiber length ( y) y = 0.4346x+53.957(R~2=0.6017). It is feasible to use hyperspectral data inversion boll biochemistry / quality indicators to monitor the cotton fiber quality indicators.