High precision population spatial data plays an extremely important role in urban resource allocation and planning. Aiming at the problem that it is difficult to meet the needs of regional large-scale research with the currently published population data sets. In this study, night light data, POI data, land cover data and topographic data were selected to establish a population-based factor feature database, and the LightGBM model based on Bayesian optimization was used to model township (street) population in Henan Province in 2020, and the spatial distribution of population density in Henan province with a resolution of 200 meters was retrieved. The accuracy of the seventh population census data in 2020 was verified with the model results. The coefficient of determination R² of the research results was 0.932, the average absolute error was 0.134, and the mean square error was 0.034, which showed high accuracy, proving that the data set obtained in this study could better reflect the population density distribution of Henan Province in 2020. It can be used for higher precision population spatial analysis.