Geographical Supervision Correction for Remote Sensing Representation Learning

Global land cover (GLC) products can be utilized to provide geographical supervision for remote sensing representation learning, which has significantly improved downstream tasks' performance and decreased the demand of manual annotations. However, the time differences between remote sensing images and GLC products may introduce deviations in geographical supervision. In this article, we propose a geographical supervision correction (GeCo) method for remote sensing representation learning. Deviated geographical supervision generated by GLC products can be corrected adaptively using the correction matrix during network pretraining and joint optimization process is designed to simultaneously update the correction matrix and network parameters. In addition, we identify prior knowledge on geographical supervision to guide representation learning and restrict the correction process. The prior knowledge named minor changes implies that the geographical supervision may not change significantly, whereas the prior knowledge named spatial aggregation implies that land covers are aggregated in their spatial distribution. According to the prior knowledge, corresponding regularization terms are proposed to prevent abrupt changes in the geographical supervision correction process and excessive smoothing of network outputs, thereby ensuring the adaptive correction process's correctness. Experimental results demonstrate that our proposed method outperforms random initialization, ImageNet pretraining, and other representation learning methods on a variety of downstream tasks. In particular, when compared to the method that learns representations directly from deviated geographical supervision, it is proven that our method can eliminate the influence of deviations and further improve the effect of representation learning.

Automated summary

A geographical supervision correction (GeCo) method is proposed for remote sensing representation learning, which adapts to correct deviated geographical supervision generated by GLC products using a correction matrix during network pretraining. Prior knowledge is identified to guide representation learning and restrict the correction process, with corresponding regularization terms proposed to prevent abrupt changes and excessive smoothing of network outputs. Experimental results demonstrate that the method outperforms random initialization, ImageNet pretraining, and other representation learning methods, and can eliminate the influence of deviations to improve the effect of representation learning.