Study On The Spectral Features Of Camellia Sinensis Based On The Diversities Of Endmember

Typical spectrum data represented most kinds of surface features is the basic theory of remote sensing. It is the important fundamental information for the design of sensor or applications of remote sensing also. Researching the rules of the spectrum variation of tea tree (Camellia sinensis) in different spatio-temporal scales and the relationship between spectrums with biophysical parameters, can provide key technical support of endmember extraction and pixel unmixing for the aim of tea plantation management.In this paper, endmembers of tea tree was used as an entry point of study. The spectral differences of various objicts were covered all researches. Researches focus on the spectral characteristics of tea tree at different spatial and temporal scale, and the spectrum properties influenced by biophysical parameters. At the spatial scales, the spectrum properties of leaves, canopy and tea plantation were studied. At the temporal scale, diurnal and seasonal variations of tea tree canopy spectrum were studied. At the aspects of biophysical parameters, three biophysical parameters influenced the spectrum of tea leaves or of tea tree were studied. The three biophysical parameters are water content, fresh leaves in canopy and leaf area index (LAI). The relations between water contents and the spectra in the tea leaves were studied. Models for inversting water contents were built at the same time. Correlation analyses and modeling was done between the unit area weight of fresh leaves in canopy and canopy spectrum. In addition, the spectrum of tree canopy influenced by canopy structure was explored and the retrieval model of leaf area index (LAI) was built.On the leaf level, there is a clear difference between the new fresh tea leaves and the old one.There are no obvious differences in new leaves they grew at different part of branch. Study has found that120cm (25degree FOV) or310cm (10degree FOV) height was suitable for spectra measuring for ridge planting vegetable link tea tree. The spectral reflectances are different at two sides of the ridge of tea tree due to differences in irradiance. That different value can be corrected by "Cosine Correction". The spectral reflectance in tilt ridge can be corrected to the one of Horizontal ridge position. There are four kinds of endmembers in tea plantation. Four-component synthesis method can be used for the compound of the simulated mixed pixel of tea plantation in hillsides. The accuracy of synthesis method was above85%.The diurnal variations of spectral reflectance of tea canopy were low in the morning, high at noon and low in the afternoon again. Vegetation Indexs (VI) are insensitive to the spectrum measuring time and to ridge position. So, VI can be used for inversions of biochemical component and of other parameters. The rules of seasonal variation of tea canopy spectra are:green peak and red valley was no significant change in position, red edge position remains unchanged in the early growth stage and moves to longer wavelengths near maturity. Red edge normalized index, normalized of greenness index and photochemical reflectance index is increases with new leave growing. After comparing the spectral characteristics of original fresh tea leaves and the one of complete dehydration (drying), it was found that the reflectance values of drying leaf increased, and was evident particularly in the1320-2500nm region. For drying leaf, yellow peak appeared, green peak position shifted clearly to longer wavelengths, red valley position to shortwave direction displacement, red edge to shorter wavelength. Water absorption valley disappeared or became obvious. In the visible part, the correlation of leaf water content expressed by weight moisture content (FMC) with spectrum was high than the one of leaf water content expressed by equivalent water thickness (EWT). Whereas in the near infrared to shortwave infrared region (760-2500nm), FMC has a few high correlation bands, while EWT has not only a large number of high correlation bands but also high correlation coefficient, especially after1370nm. On the FMC estimation, using Simple Radio Water Index (SRWI) calculated by original spectrum data was better than other indices in the inversion of leaves FMC. Inversion error ranged within10%and has a overall accuracy of96%. Many spectral indexes can be used for modeling the EWT. MSI, NDWI1640(normalized spectral index using the reflectance of860nm and1640nm) and RADIO1200(three-band index,1200nm as the center of bands) did best especially. The EWT linear relationship model were:YEWT=-0.0507XMSI+0.0553, YEWT=0.1112XNDWI1640+0.0081, YEWT=-0.2633XRADIO1200+0.2673. Estimated EWT and measured one was close, the model R2was above0.7, accuracy was over93%. Results accuracy of estimated value using three models above was better than one of Model built by a single-band spectrum.The ratio Index (R522/R675) was calculated using the spectral reflectance of522nm and675nm (before picked). Its linear model (Y=930.25X-2079, X is the ratio index, Y is fresh leaf weight, g/m2) can be use for estimation of canopy fresh tea weight.Correlations of LAI with spectral parameters and modeling effects are:spectral characteristic value method, the first derivative spectra Law> Spectral Index> Original spectrometry. A new model (Water Chlorophyll model) for LAI inversion was found in this paper. The model used the radio (R1490/Rg) of spectral reflectance in1490nm (in water absorption band peak position) and spectral reflectance in green peak. Quadratic polynomial model (Y=0.8704X2-6.0051X+10.818) and linear model (Y=-3.2132X+8.6808) cound be used for LAI inversion and was better than other model. Deviation of retrieved LAI was only0.65, accuracy of model was greater than80%.