1. College of Electrical and Information, Heilongjiang Bayi Agricultural University, Daqing, Heilongjiang 163319, China; 2. College of Electrical and Information, Northeast Agricultural University, Harbin, Heilongjiang 150030, China; 3. College of Agriculture, Heilongjiang Bayi Agricultural University, Daqing, Heilongjiang 163319, China
Abstract：The extraction techniques of multiple spectral characteristic wavelength band were applied, and the nearinfrared spectral characteristics of soybean canopy during the whole growth period were investigated. An estimation model of SPAD value for soybean canopy was proposed based on the multidimensional spectral characteristic wavelength extraction. For the original nearinfrared spectral curve of soybean canopy, the multiple scattering correction preprocessing and the partial least squares regression modeling were preferred. Preprocessed with multiple scattering corrections, the competitive adaptive reweighted sampling algorithm (CARS), the principal component analysis and the successive projections algorithm were applied to extract the spectral characteristic wavelengths of soybean canopy, which were respective 22, 51 and 12. Based on the characteristic wavelengths, the partial least squares regression and the multiple linear regression (MLR) methods were used to establish the estimation models of soybean canopy SPAD values. The results show that the CARSMLR model has good experimental effect, and the root mean square errors of CARSMLR model for corrected and predicted samples are respective 5.67 and 5.94 with the average error of about 5.81.
