In this paper, a lightweight navigation landmarks detection method based on KDS R-CNN is proposed. Firstly, aerial videos of navigation landmarks are acquired by UAVs within the cruising range, and a visual navigation dataset is produced by data augmentation and automatic labeling of images. Secondly, a lightweight KDS R-CNN network model is constructed, which significantly reduces the number of parameters and computation of the model. Then a knowledge distillation strategy is used to train the lightweight KDS R-CNN network so as to improve its detection accuracy. Finally, the lightweight KDS R-CNN network is deployed on an embedded platform and validated by inference using the visual navigation dataset. Experimental results show that the algorithm proposed in this paper is easier to be deployed in embedded platforms. In addition, the trained KDS R-CNN model can accurately detect the navigation landmarks, which lays the foundation for the visual navigation of the aircraft.