Abstract:Precise prediction of chlorophyll content in different vertical positions of tomato canopy is an important indicator for timely prevention and control of tomato diseases and insect pests, precise fertilization and irrigation. UAV can quickly and efficiently obtain crop canopy spectral information, which facilitates agricultural production. Aiming to predict the soil and plant analysis development (SPAD) values of different vertical positions of tomato canopy by using multispectral remote sensing images of UAV. Firstly, a UAV equipped with a multispectral camera (Sequoia) was used to obtain multispectral images of the tomato blooming and fruit setting stage, fruiting early stage and fruiting late stage. At the same time, SPAD-502Plus chlorophyll meter was used to measure the SPAD values of the upper, middle, lower and the whole canopy of tomato. The SPAD values of the three growth periods of tomato showed that the SPAD values of the upper leaves of tomato canopy were higher than those of the middle and lower leaves in the fruit setting stage, and the SPAD values of the middle leaves of tomato canopy were higher than those of the upper and lower leaves in the fruiting stage. Secondly, RTK was used to record the location of sampling points to establish region of interest (RoI) and extract the reflectivity of each band in RoI. Vegetation index was calculated according to the reflectance data. The correlation and sensitivity between SPAD values and vegetation index of tomato upper, middle, lower and the whole canopy were analyzed. Finally, the best vegetation index was selected and the prediction model of SPAD value was established. The study results were as follows: the correlation degree and linear sensitivity of SPAD values and vegetation index of the upper canopy leaves were better than those of the middle and lower canopy leaves. In the same prediction model, R2 value of the upper and the whole canopy prediction model was higher than that of the middle and the lower canopy, so it was difficult to accurately predict the chlorophyll content of the lower canopy only by using the canopy spectrum. The R2 value of support vector machine (SVR) model in the upper, middle and lower layers of canopy and the whole canopy was higher than that of partial least squares (PLS) and BP neural network model. The research result provided a method basis for UAV to accurately predict tomato canopy chlorophyll.
