An Attention Cascade Global–Local Network for Remote Sensing Scene Classification

Remote sensing image scene classification is an important task of remote interpretation, which has recently been well addressed by the convolutional neural network owing to its powerful learning ability. However, due multiple types geographical information and redundant background images, most CNN-based methods, especially those based on a single CNN model ignoring combination global local features, exhibit limited performance accurate classification. To compensate for such insufficiency, we propose new dual-model deep feature fusion method attention cascade global–local (ACGLNet). Specifically, use two popular CNNs as extractors extract complementary multiscale features from input image. Considering characteristics proposed ACGLNet filters low-level through spatial mechanism, followed locally attended are fused with high-level features. Then, bilinear employed produce representation dual model, finally fed classifier. Through extensive experiments four public datasets, including UCM, AID, PatternNet, OPTIMAL-31, demonstrate feasibility superiority over state-of-the-art methods.