Plant height (H) and above-ground biomass (AGB) are important parameters for monitoring growth and evaluating yield of crop. It is significant for agricultural precision fertilization management to acquire plant H and AGB information of potato quickly and accurately.Hyperspectral images, measured plant height (H), measured above-ground biomass (AGB) and three-dimensional information of ground control point (GCP) were obtained respectively at the budding stage, tuber formingstage, tuber growth period, starch accumulation period and maturity period of potato.Firstly, the digital surface model (DSM) of test field was generated based on the unmanned aerial vehicle(UAV)hyperspectral gray images combined with GCP,and the potato plant height(Hdsm ) was extracted by using DSM.Then correlation analysis of the potato AGB with the original canopy spectrum and hyperspectral indexes was performed, and the optimal spectral parameters and top 10 spectral parameters were selected, and the univariate model was constructed by exponential regression (ER) with plant height and optimal spectral parameters, respectively. Finally,multiple linear regression (MLR),partial least square regression (PLSR)and random forest (RF)were used to construct and compare the AGB estimation model at different growth periods to select the optimal model. The results showed that the Hdsm extracted from the UAV images was highly fitted with the measured plant height (H) (R 2 =0.84); in the univariate model, the verification accuracy of AGB estimated by ER in each growth period was higher than that of corresponding modeling accuracy, in which the effect of the model was in the order of optimal spectral parameters, Hdsm and H, and the estimation accuracy of CIrededge was the highest (R 2 =0.45) in the tuber growth period; in the multivariable model,three methods were used to construct AGB estimation model for each growth period, and the model with spectral index added to Hdsm had higher accuracy in each method. The effect of AGB model with spectral index and Hdsm of MLR (R 2 was 0.64, 0.70, 0.79, 0.68 and 0.63) was better than that of PLSR (R 2 was 0.62, 0.68, 0.75, 0.67 and 0.60) and RF (R 2 was 0.56, 0.61, 0.67, 0.63 and 0.53) in each growth period. The potato AGB was mapped by using the MLR model, and the AGB distribution was consistent with the actual growth situation in the five growth stages.The MLR model integrated with Hdsm can be used to estimate the potato AGB in a large area, which provided technical support for the quantitative research of precision agriculture.