STA-GAN: A Spatio-Temporal Attention Generative Adversarial Network for Missing Value Imputation in Satellite Data

Satellite data is of high importance for ocean environment monitoring and protection. However, due to the missing values in satellite data, caused by various force majeure factors such as cloud cover, bad weather sensor failure, quality reduced greatly, which hinders applications practice. Therefore, a variety methods have been proposed conduct imputation improve its quality. these cannot well learn short-term temporal dependence dynamic spatial resulting performance when rate large. To address this issue, we propose Spatio-Temporal Attention Generative Adversarial Network (STA-GAN) value data. First, develop (STA) mechanism based on Graph (GAT) features capturing both Then, learned from STA are fused enrich spatio-temporal information training generator discriminator STA-GAN. Finally, use generated trained STA-GAN fill Experimental results real datasets show that largely outperforms baseline methods, especially filling with large rates.