| Quantitative monitoring of crop biochemical status is an important research field in the agricultural remote sensing. Non-destructive, fast, accurate and real time estimation the biochemical status is one of key techniques in the development of precision agriculture. The primary objective of this study was to explore the optimum wavebands, sensitive spectral indices and accurate, general applicative quantitative models for estimating the nitrogen, nicotine and pigment indices of flue-cured tobacco through systematic extraction of hyperspectral information at leaf level. The field experiments was based on two consecutive years under different light quliaty, light intensity and lighting time, taking Yunyan87as experimental variety. The anticipated results would provide utilizing space-borne remote sensing data, and thus assist in effective techniques for real time estmation and precise diagnosis of flue-cured tobacco nitrogen, nicotine and pigment.Firstly, the changes of nitrogen, nicotine and pigments during the growth stage, and the quality of baked tobacco leaf under different light conditions were researched. The result showed that the total content of organic acid would increase and the total content of neutral aroma components would decrease under the changed light quality, reduced light intensity or lower lighting time. The content of chlorophyll a, chlorophyll a+b and carotenoid were added compared with the control under different light quality treatmens during the middle and late growth stage. The increase proportion of yellow light encouraged to improve the nicotine content and reduce the nitrogen content. The increase proportion of red light improved the nitrogen and potassium content, and lessened the total sugar, reducing sugar, nicotine, starch and petroleum ether extract content. The nitrogen content was increased and the nicotine content in the baked tobacco leaves was decreased under the shading treatments. Moderate and heavy shading(S2, S3) improved the nitrogen content in the baked tobacco leaves, and the potassium content in the middle leaves, and reduced the starch conten in the middle leaves. Compared with the control, the lighting time treatments all promoted the increase of nicotine, nitrogen and pigments content. While T3reduced the reducing sugar, total sugar, starch and prtroleum ether extract content obviously, and improved the improved the nitrogen and potassium content in baked tobacco leaves. These results can provide the theoretical guide to improve the tobacco leaf quality in field production.Further analysis was conducted on the relationship between nitrogen, nicotine, pigment and the spectral reflectance. Firstly, a redced precise sampling method was adopted for comprehensive analysis of the quantitative relationship of nitrogen content to NDVI (Normalized different vegetation index, NDVI) and SRI(Spectral ratio index, SRI) composed of any two wavebands with original reflectance within the spectral range of350-2500nm. Then the best sensitive spectrum index and the sensitive wavebands with different bandwidths was selected. Finally, the monitoring equations were constructed for nitrogen content in flue-cured tobacco leaves. The results revealed that the best spectrum indices for nitrogen were NDVI (1970,650) and SRI (590,1980), and the sensitive wavebands were590nm,650nm,1970nm,1980nm. The range of sensitive wavebands were concentrated in the visible and short-infrared light ranges. The tobacco leaf nitrogen content monitoring models developed from NDVI (1970,650) and SRI (590,1980) gave high estimation accuracy, and the BP neural network had the best estimation.The results could be used to predict the nitrogen content in flue-cured tobacco under different light quality.By elucidating the change patterns of the single leaf hyperspectrum and nicotine content at different growth stages under different light intensity, analysis was made on the accuracy and stability of monitoring leaf nicotine status with the single hyperspectral parameters including the new spectral indices. The results showed that the difference of spectral reflectance between different light intensity treatments was the most abvious in the vigorous growth stage, and the spectral reflectance of flue-cured tobacco leaves decreased by the shading degree increased. The new spectrum indices for nicotine were NDVI (2150,610) and SRI(450,500), and the sensitive wavebands were450nm,500nm,610nm and2150nm. The range of sensitive wavebands were concentrated in the visible and short-infrared light ranges. The R2of the tobacco leaf nicotine content monitoring models(linear models of NDVI(2150,610) and SRI(450,500), SMLR, BP neural network) developed from NDVI(2150,610) and SRI(450,500) was0.796,0.810,0.842and0.968, respectively, and RMSE was0.784,0.958,0.883and0.109, respectively. The results indicated that hyperspectral remote sensing can be used for quick and accurate monitoring of the leaf nicotine content and shading status of flue-cured tobacco.The quantitative relationships were analyzed between the leaf chlorophyll a content and the new spectral indices of NDVI, SRI, DSI(Difference spectral index) composed of any two wavebands within the spectral range of350-2500nm. From the derived core bands and sensitive parameters, the monitoring equations were constrcted for leaf chlorophyll a content in flue-cured tobacco. The results showed that the sensitive wavebands of chlorophyll a content were mostly located in the visible and short-infrared light regions. The best spectrum indices for chlorophyll a status were NDVI (470,440), SRI (580,2090) and DSI (440,470), and the sensitive wavebands were440nm,470nm,580nm and2090nm. The leaf chlorophyll a content linear and BP neural network monitoring models developed from NDVI (470,440), SRI (580,2090) and DSI (440,470) gave high estimation accuracy. The R2of the linear models of NDVI (470,440), SRI (580,2090), DSI(440,470) and BP neural network was0.877,0.862,0.877and0.952, respectively. And the RMSE was0.877,0.301,0.472and0.065, respectively. The BP neural network generated the best estimation. The results could provide a new technique and sensitive wavebands rapid and non-destructive for estimation the chlorophyll a content in the flue-cured tobacco. |
PDF Full Text Request