The invention discloses an unmanned aerial vehicle autonomous flight control method and system. The method comprises the following steps: firstly creating a generation network and a classification anddiscrimination network; determining a false sample according to a training set A and the generation network; and then inputting the false sample and the training set into the classification and discrimination network to train, and acquiring an output parameter of the classification and discrimination network wherein a network output result is one-dimension more than a scene instruction mark vector, and the one dimension feature is used for classifying true and false samples, and the false sample with instruction mark and the training set A with the instruction mark can be used as the real sample to input into the generation network to acquire an output parameter of the classification and discrimination network in the process of training the generation network, and then the accuracy of theclassification and discrimination network parameter determination is improved, and the autonomous flight control precision is further improved.