Hongxiang Ma

The lack of remote sensing images and poor quality limit the performance improvement of follow-up research such as remote sensing interpretation. In this article, a generative adversarial network (GAN) is proposed for data augmentation of remote sensing images abstracted from Jiangxi and Anhui Provinces in China, i.e., deeply supervised GAN (D-sGAN). D-sGAN can generate high-quality images that are rich in changes, greatly shorten the generation time, and provide data support for applications such as semantic interpretation of remote sensing images. First, to modulate the layer activations, a downsampling scheme is designed based on the segmentation map. Then, the architecture of the generator is Unet++ with the proposed downsampling module. Next, the generator of this net is deeply supervised by the discriminator using deep convolutional neural network. This article further proved that the proposed downsampling module and the dense connection characteristics of UNet++ are significantly beneficial to the retention of semantic information of remote sensing images. Numerical results demonstrated that the images generated by D-sGAN could be used to improve accuracy of the segmentation network, with the faster generation speed compared to the CoGAN, SimGAN, and CycleGAN models. Furthermore, the remote sensing data generated by the model helped the interpretation network to increase the accuracy by 9%, meeting actual generation requirements.

Since remote sensing images are difficult to obtain and need to go through a complicated administrative procedure for use in China, it cannot meet the requirement of huge training samples for Waterside Change Detection based on deep learning. Recently, data augmentation has become an effective method to address the issue of an absence of training samples. Therefore, an improved Generative Adversarial Network (GAN), i.e., BTD-sGAN (Text-based Deeply-supervised GAN), is proposed to generate training samples for remote sensing images of Anhui Province, China. The principal structure of our model is based on Deeply-supervised GAN(D-sGAN), and D-sGAN is improved from the point of the diversity of the generated samples. First, the network takes Perlin Noise, image segmentation graph, and encoded text vector as input, in which the size of image segmentation graph is adjusted to 128 × 128 to facilitate fusion with the text vector. Then, to improve the diversity of the generated images, the text vector is used to modify the semantic loss of the downsampled text. Finally, to balance the time and quality of image generation, only a two-layer Unet++ structure is used to generate the image. Herein, “Inception Score”, “Human Rank”, and “Inference Time” are used to evaluate the performance of BTD-sGAN, StackGAN++, and GAN-INT-CLS. At the same time, to verify the diversity of the remote sensing images generated by BTD-sGAN, this paper compares the results when the generated images are sent to the remote sensing interpretation network and when the generated images are not added; the results show that the generated image can improve the precision of soil-moving detection by 5%, which proves the effectiveness of the proposed model.