Abstract: Taking the bergamot in Nongyuan Village, Haohuahong Town, Huishui County, Guizhou Province as the research object, through acquiring the multi spectral images of the unmanned aerial vehicle （UAV） in the bergamot planting base, the vegetation index were calculated, the correlation between different vegetation indexes and SPAD values were analyzed, and the input variables of the model was screened out. Using partial least squares （PLS） and random forest （RF） to build SPAD estimation models, the SPAD values of green and yellow leaves of bergamot were estimated. The results showed that, besides CIre, NDVI, GNDVI, RVI and SPAD values had a strong correlation. In the PLS estimation model of green leaves and yellow leaves, the validation set R² were 0.814 and 0.660 respectively, and the RMSE were 1.12 and 1.16 respectively. The SPAD value estimation effect of green leaves was higher than that of yellow leaves. In RF model, R² of validation set and model building was lower than that of PLS model. Therefore, based on multispectral images, the SPAD estimation model of bergamot leaves established by PLS method was better than the RF model, which could realize the rapid estimation of SPAD values of bergamot leaves.

Key words: leaves of bergamot（Sechium edule）, UAV, multispectrum, SPAD