AbstractThe net photosynthetic rate (Pn) is one of the important indicators to measure photosynthetic capacity of crops. Therefore, it is critically important to find real‐time methods for accurately estimating Pn of winter wheat (Triticum aestivum L.). This information could provide guidance on the management of waterlogging stress. To explore the optimal monitoring method for Pn of winter wheat under waterlogging stress, the correlations between Pn and 16 characteristic image indices were analyzed in irrigated and drained microplot experiments. Then, based on the indices values, Pn was estimated using the multiple linear regression (MLR), support vector machine (SVM), backpropagation neural network (BP), and random forest (RF) models, which were constructed based on the optimal monitoring image indices. Water logging when compared to no waterlogged wheat had similar Pn values <6 days. After 12 days, waterlogged wheat plants had a lower Pn value than in no waterlogged plants. All indices were correlated with the Pn (p < 0.05), and Pn estimation accuracy was lower at the winter wheat flowering and complete ripeness periods than at the milky and waxy ripe maturity periods. Based on the results of the models tested, the RF model had higher R2 (0.904) values than the other models. These findings suggest that machine learning models could be used to accurately predict Pn, and the random forest algorithm was the best.