MapsNet: Multi-level feature constraint and fusion network for change detection

• DOI: 10.1016/j.jag.2022.102676
• Publication Date: 2022-03-04T01:29:39Z
• Authors: Jianping Pan; Wei Cui; Xinyong An; Xiao Huang; Hanchao Zhang; Sihang Zhang; Ruiqian Zhang; Xin Li; Weihua Cheng; Yong Hu
• Keywords: Robustness; Feature (linguistics)

Nowadays, the tidal waves of deep convolution have promoted proliferation learning change detection (CD) methods. However, challenges still remain as most algorithms tend to poor detections small targets, unsmooth edges, and incomplete internal regions, largely because a lack effective features, context information, feature fusion. In this paper, multi-attention feature-constrained pixel-shuffle image fusion network (MapsNet) is proposed address in CD tasks. We first employ two-stream fully convolutional for extraction, which adaptively constrained by module (MAM). The capability MAM further enhanced introduction novel attention module, i.e., CBAM-s. addition, we propose (PSIFN) aggregate multi-level contextual information implement complete map reconstruction. conducted experimental results confirm that MapsNet demonstrates better effectiveness robustness complex changing scenes compared selected state-of-the-art Finally, these strategies designed article high practicability transferability.