Research on Remote Sensing Sample Expansion Technology Based on Generative Adversarial Network

In some specific application scenarios, it is difficult to obtain enough public datasets for remote sensing images, which limits the application of object detection algorithms in the field of remote sensing images. In order to solve this problem, this paper proposes Res-WGAN: an image sample expansion technology based on generative adversarial network. Res-WGAN combines variational autoencoder and WGAN network, so that the generated images are closer to the original sample distribution. By introducing the residual block, the depth of the network is deepened and the resolution of the generated remote sensing images is improved. The experimental results show that the proposed Res-WGAN can make the generated images more clear and diverse. Furthermore, the generated remote sensing images are used to train the Faster R-CNN model to verify the effectiveness of sample expansion. The experimental results show that the expanded sample set can effectively improve the accuracy of object detection.