MDCGA-Net: Multiscale Direction Context-Aware Network With Global Attention for Building Extraction From Remote Sensing Images
Penghui Niu, Junhua Gu, Yajuan Zhang, Ping Zhang, Taotao Cai, Wenjia Xu, Jungong Han
Abstract
Building extraction from remote sensing images (RSIs) requires exploring multi-scale boundary detailed information and extracting it completely, which is challenging but indispensable. However, existing solutions tend to augment feature information solely through multi-scale fusion and apply attention mechanisms to focus on feature relationships within a single layer while ignoring the multi-scale information, which affects segmentation results. Therefore, enhancing the capability of the network to adaptively capture multi-scale information and capture the global relationship of features remains a pivotal challenge in overcoming the aforementioned hurdles. To address the preceding challenge, we propose a <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><u>M</u>ulti-scale <u>D</u>irection <u>C</u>ontext-aware network with <u>G</u>lobal <u>A</u>ttention</i> (MDCGA-Net), employing a classic encoder-decoder architecture enhanced with direction information and global attention flow. Specifically, in the encoder part, the multi-scale layer (MSL) is used to extract contextual information from the inter-layer. Additionally, the multi-scale direction context-aware module (MDCM) is adopted to adaptively acquire multi-scale information. In the decoder part, we propose a global attention gate module (GAGM) to capture discriminative features. Furthermore, we construct an operation of attention feature flow to obtain the global relationship among the different features with long-range dependencies, which guarantees the integrity of results. Finally, we have performed comprehensive experiments on three public datasets to showcase the efficacy and efficiency of MDCGA-Net in building extraction.