Hyperspectral imaging technology for detection of moisture content of tomato leaves

Abstract

Hyperspectral imaging technology for detection of moisture content of crops takes into account both the internal information and external features. It improves the comprehensiveness and reliability of detection. A hyperspectral imaging system is developed to perform acquisition of hyperspectral imaging data. The adaptive band selection is adopted to select the optimal characteristic wavelength from lots of data, and the optimal wavelength is 1420nm. The images of all samples at 1420nm are segmented, reversed and operated, and then the target images are obtained. The mean value and standard deviation of gray scale are extracted as grey features, and the mean value and standard deviation of energy, entropy, geometrical moment of inertia, correlation as texture features. The optimal feature subset is selected by GA-PLSR, and then the partial least-squares regression model is established. The correlation coefficient between the predict value and the real value is 0.902. It is higher obviously than the prediction models based on grey features or texture features.