Remote Sensing Image Scene Classification Based on an Enhanced Attention Module

In remote sensing image field interpretation, categorizing various satellite remote sensing scenes is a crucial subtask. Convolutional neural networks (CNNs), which have been developed recently, have improved classification techniques remote sensing scene. However, because remote-sensing image scenes often contain a large number of small objects and a complicated background, using recognition algorithms based on CNNs can be difficult. This work presents the construction of an enhanced attention module (EAM) that improves the feature extraction and generalization capabilities of deep neural networks, allowing them to learn more discriminative features. The experimental results demonstrate that the implemented model can successfully improve the accuracy of scene classification for remote sensing images, and can learn more discriminative features than the existing methods. When compared to existing methods like patch-to-region framework, multi-scale attention feature extraction block (MSAFEB), implemented method achieved high accuracy values of 99.50%, 99.62%, and 99.58% using three datasets like AID, WBDS, and NWPU datasets.