In this paper the in-situ experiment data are collected from the Spectral Library to estimate the possible values of the primary structural parameters in each cotton growing season by statistics. Based on these values...In this paper the in-situ experiment data are collected from the Spectral Library to estimate the possible values of the primary structural parameters in each cotton growing season by statistics. Based on these values the spectra of the cotton major growing seasons were simulated and analyzed using the canopy reflectance model, SAILH. In this way we also simulated the cotton pixel spectra corresponding to CBERS-02 CCD and took them as the reference spectra for spectra fitting. Then No.143 Regiment of No.8 Agricultural Division of Xinjiang Production and Construction Corps (XPCC) was chosen as the study area, and two spectra fitting methods, Mahalanobis distance and spectra angle, were used to identify cotton pixel from CBERS-02 CCD image of the study area. At last we analyzed the effect of the calibration coefficients error on cotton identifying. The results showed that within 4% error the spectral angle classifier can still have a good performance.展开更多
文摘In this paper the in-situ experiment data are collected from the Spectral Library to estimate the possible values of the primary structural parameters in each cotton growing season by statistics. Based on these values the spectra of the cotton major growing seasons were simulated and analyzed using the canopy reflectance model, SAILH. In this way we also simulated the cotton pixel spectra corresponding to CBERS-02 CCD and took them as the reference spectra for spectra fitting. Then No.143 Regiment of No.8 Agricultural Division of Xinjiang Production and Construction Corps (XPCC) was chosen as the study area, and two spectra fitting methods, Mahalanobis distance and spectra angle, were used to identify cotton pixel from CBERS-02 CCD image of the study area. At last we analyzed the effect of the calibration coefficients error on cotton identifying. The results showed that within 4% error the spectral angle classifier can still have a good performance.