China has a huge and intensive open-field vegetable production system. However, serious issues such as low water and nitrogen use efficiencies, as well as excessive fertilizer application, limit the sustainability of the system. To improve the efficiency of production and enhance accurate fertilization in large-scale vegetable cultivation systems, this study was conducted with open-field lettuce. Three treatments of no nitrogen (N0), low nitrogen (N1) and high nitrogen (N2) were established. An unmanned aerial vehicle (UAV) equipped with a multi-spectral camera was used to establish correlations between three multi-spectral vegetation indices (NDVI, RVI, and SAVI) and lettuce chlorophyll content, biomass, crop nitrogen uptake, and total nitrogen content. Models to predict total nitrogen content for single growth stage and multiple growth stages were developed. The results showed that: (1) during rosette and heading stages, NDVI, RVI and SAVI values increased with the amount of applied nitrogen, but that during harvest stage, maximum values occurred with the N1 treatment; (2) NDVI showed a significant correlation with lettuce yield, nitrogen uptake and chlorophyll content during heading stage; and the total nitrogen content of lettuce was significantly correlated with chlorophyll content at p<0.01 level, with a correlation coefficient (R) of 0.51. When considering multiple growth stages together, NDVI values showed a significant correlation with lettuce yield, chlorophyll content, nitrogen uptake, and total nitrogen content at p<0.001 level, with correlation coefficients of 0.85, 0.82, 0.81, and 0.71, respectively. (3) Relationships of exponential, linear, logarithmic and power functions were fitted to the corresponding datasets, and the best prediction model of total nitrogen content for lettuce (N%=16.52ln(NDVI)+73.514) was established in the multiple growth stages. Using the lettuce total nitrogen content prediction model to obtain modeled values of total nitrogen content for the area of a commercial production field on the same farm, the average relative error was 3.2%, RMSE=0.556 6, NRMSE=0.010 8, showing accurate estimation of total nitrogen content. The results show that the model had good accuracy and that it is feasible to diagnose vegetable nitrogen content using unmanned aerial vehicle multi-spectral remote sensing.