%A DING Yong-jun;ZHANG Jing-jing;SUN Hong;LI Xiu-hua %T Sensitive Bands Extraction and Prediction Model of Tomato Chlorophyll in Glass Greenhouse %0 Journal Article %D 2017 %J SPECTROSCOPY AND SPECTRAL ANALYSIS %R 10.3964/j.issn.1000-0593(2017)01-0194-06 %P 194-199 %V 37 %N 01 %U {https://www.gpxygpfx.com/CN/abstract/article_8907.shtml} %8 2017-01-01 %X In order to predict the content of chlorophyll in tomato rapidly and accurately, this study, with spectrum technology, focused on the extraction of sensitive spectral bands of tomato chlorophyll in glass greenhouse environment and created an effective estimation model. During the period of cultivating tomatoes, leaf spectra were measured with an ASD FieldSpec HH spectrophotometer and chlorophyll content was measured with Type 752 UV-Vis spectrophotometer. Based on the original spectra, absorbance spectra, first derivative spectra and continuum removal spectra, spectral data was preprocessed, in which the effectiveness of spectral features of chlorophyll content of tomato was highlighted and spectral response characteristics of the absorbance spectra in the visible part was enhanced. It was shown that both the continuum removal spectra and the first derivative spectra have strong blue and red absorption valley and green reflection peak. In this paper, the original spectrum, absorbance spectrum, first derivative spectrum and continuum removal spectrum were analyzed and compared, and then optimal spectral feature parameters were extracted with methods of Inter-Correlation analysis and multivariate collinearity diagnosis. Sensitive bands from original spectrum are 639, 672, 696, 750 and 768 nm. Sensitive bands from absorbance spectrum are 638, 663, 750 and 763 nm. Sensitive bands from first derivative spectrum are 516, 559 and 778 nm. Sensitive bands from continuum removal spectrum are 436, 564, 591, 612, 635, 683 and 760 nm. The stepwise multiple regressions were used to develop the prediction models of the chlorophyll content of tomato leaf. The result showed that the prediction model, which used the values from continuum removal spectrum at 436, 564, 591, 612, 635, 683, 760 nm as input variables, had the best predictive ability. The calibration R-Square was 0.88 and the validation R-Square was 0.82.