Water moisture diagnosis of tomato canopy based on multi-information fusion

In order to accurately, rapidly and stably determine the moisture content of tomatoes and fully detect water stress degree, the multi-information fusion of the reflection spectrum, multi-spectral images, canopy temperature, environmental temperature and humidity were used to judge the water stress. After removing the noise wavelengths and correlation analysis, 950-1 080, 1 170-1 300, 1 370-1 500, 1 600-1 730 and 1 860-1 980 nm wave bands were selected. Bands of 1026, 1247, 1431, 1640 and 1910 nm were as the features using the stepwise regression. The 3×3 window median filter and the Otsu segmented were applied by IR and G images，finally the average gray was calculated. Combined with canopy temperature, environmental temperature and humidity, the CWSI was established. The five characteristic wavelengths, IR and G images average gray and CWSI as multi-information fusions parameters, the analysis and verification results obtained by PLS-ANN were correlation coefficient 0.9364, the root mean square error 10.6713, mean error 7.6714%, fitting equation slope 0.9615. It showed that the evaluations of the multi-information fusion model were better than those of the model represented by single sensor.