Reconstruction of Missing Information in Satellite Images Based on Convolutional Neural Networks

Due to the failure and damage of satellite detectors, the acquired remote sensing data often suffer from information loss, which makes the quantitative application of satellite data much less effective. In this paper, a deep convolutional neural network based satellite remote sensing image information reconstruction method is proposed to reconstruct the missing information, using spectral-spatial-temporal information as supplementary information. The model parameters are determined by experiments in the model building process. Finally, the model is applied for the actual missing data of Fengyun-3 satellite, and the results of data experiments show that the proposed model performs well in the missing information reconstruction task.