Monitoring The Main Diseases In Southern Plantation Based On The Ground Of Hyperspectral Remote Sensing

Using three main plantations that Chinese fir, Pinus massoniana and oil camellia in different physical levels of health as objects the chlorophyll content and the reflectivity of canopies of all the above was measured by hand-holding field spectra radiometer made in ASD company from USA by field survey of Chinese fir anthracnose, Pinus massoniana needle blight and oil camellia anthracnose. The hyper-spectral derivative data combined with moving average filter was pretreated, then the red edge characteristic parameters were extracted analysing correlation among the original spectra, the derivative spectra, disease and chlorophyll content. The inversion models of the deasase and the chlorophyll content were built by analysing the relationship among the red edge characteristic parameters, the deasase and the chlorophyll content using curve fitting and multivariate statistical analysis, the prediction model of the chlorophyll content in the leaves of canopies of oil camellia under anthracnose stress was built by discussing exploitation of groung-based hyperspectral data combined with principal component analysis (PCA) and artificial neural network algorithm at the same time, and the prediction results using different modeling methods were evaluated comprehensively by using correlation coefficient(r) and RMS error(RMSE). The results showed:(1) The derivative spectra of Chinese fir, Pinus massoniana and oil camellia could respond to the changes of the deasase and the chlorophyll content better than the original spectra, furthermore, the important indicatory region that described the healthy condition of vegetation was from the red light to the "red edge" in the near infrared region.(2) When trees were under disease stress, the chlorophyll content decreased gradually and the" red valley" due to strong absorption of the chlorophyll content disappeared gradually in red light region with the severity of the disease aggravating. At the same time, the positions of the red edge displayed "blue shift" successively with the severity of disease aggravating, and the slope of the red edge decreased little by little.(3) The correlation among the derivative spectra of canopies of oil camellia in the red edge region, the disease and the chlorophyll content mainly concentrated in the range of670～720nm, whose sensitive bands moved about30nm to the shortwave compared with Chinese fir and Pinus massoniana; the prediction ability of stepwise regression model between disease and chlorophyll content using the red edge characteristic parameters as input variables from the point of inversion accuracy of model was worse than that of Chinese fir and Pinus massoniana.(4) The pridciton accuracy of stepwise regression model built using red edge characteristic parameters that had better indicatory function of the disease and the chlorophyll content of Chinese fir and Pinus massoniana was higher, which can be as a preferential model; the artificial neural network model of the chlorophyll content of canopies of oil camellia was built under anthracnose stress integrating correlation coefficient analysis with dimensionality reduction of PCA spectra data, whose accuracy was higher than that of simple linear regression model and multiple linear regression model, which is a preferential modeling method.This paper has discussed the methods using hyper-spectral remote sensing to monitor disease of forest through building the inversion model of the disease and the chlorophyll content of forest under disease stress, which facilitated the application of hyper-spectral remote sensing in monitoring disease of forest. Also the paper provided reference for disease of other trees and established the foundation for study on the relationship between the spectra data from aviation and aerospace remote sensing and from ground remote sensing equipment.