关键词: 模型精度, 叶绿素含量, 盆栽试验, 倒数对数, 导数变换

Abstract: In this paper，a freezing injury experiment of potted winter wheat was explored by using highlow temperature and humidity test chamber.Winter wheat freezing injury，hyperspectral reflectance and the corresponding chlorophyll content were continuously observed before and after lowtemperature treatment，and analysis their change regulation.By making the correlation analysis between the countdown logarithms，first derivative，second derivative transformations of hyperspectral and the chlorophyll content，eigenvalues for characterizing freezing stress were found，the spectral band and the index were obtained to identify and evaluate the freezing injury degree differences.The investigation results showed that，（1）the chlorophyll content and the original spectral reflectance was negatively correlated in the visible range，and they were positively correlated in the nearinfrared range.However，the chlorophyll content and the countdown logarithms of spectrum had opposite correlations.Most of the first derivative spectral band correlations were significant at the 0.01 test，but，only a small part of second derivative spectra band correlations passed the test.（2）Correlation analysis showed that the sensitive monitoring bands of freezing injury were 684.92nm reciprocal logarithmic spectrum，578.37nm at the first derivative spectrum，571.93nm at the second derivative spectrum;these bands and chlorophyll correlation coefficients passed through the 0.01 level significance test，and the correlation coefficients were maximum，among the estimation models，the one with reciprocal logarithmic spectrum as variable was the optimal.（3）Correlation analysis between chlorophyll content and hyperspectral characteristic variables showed that parabolic models with VI3（VI3=SDr/SDb）or VI5［VI5=（SDr-SDb）/（SDr+SDb）］by the blue edge area（SDb）and red edge area（SDr）calculation as variables were the best ones，the training samples and validation samples fitting accuracy test levels were the highest，so that VI3，VI5 were the sensitivity indices for winter wheat freezing injury monitoring.To conclude，this study revealed the hyperspectral characteristics of winter wheat after freezing injury.It will provide a strong support for promoting the application of hyperspectral technology in growth monitoring and yield estimation of winter wheat，and improving the remote sensing monitoring accuracy of winter wheat freezing injury.

Key words: Model accuracy, Chlorophyll content, Pot experiment, Logarithm of reciprocal, Derivative transformation