Medium-resolution multispectral satellite imagery, such as Sentinel-2 data, plays a crucial role in the field of large-scale land cover classification and crop monitoring. The accurate estimation of Chinese pepper (Zanthoxylum armatum and Zanthoxylum bungeanum) yield using Sentinel-2 data requires extensive field investigation as ground true reference data, which is challenging due to labor constraints and the limited harvest window for Chinese pepper. This study aims to overcome the challenge by utilizing unmanned aerial vehicle (UAV) technology to acquire extensive remote sensing imagery of Chinese pepper plants before harvest. By combining UAV imagery with a limited amount of field investigation data, a UAV-scale model for predicting Chinese pepper yield is developed. Subsequent integration of Sentinel-2 satellite data enables the implementation of a machine learning-based yield prediction model. The findings of the study demonstrate the commendable accuracy and reliability of the Sentinel-2-based model (with the R2 of 0.682 ± 0.01155 and RMSE of 17.720 ± 0.45646 kg/100 m2), thereby establishing a basis for the widespread implementation of automated mapping for Chinese pepper production in China. In conclusion, the study offers insights into the potential for scaling up monitoring efforts by integrating field investigations with Sentinel 2 data through the use of UAVs.