Submitted manuscript, 37.1 MB, PDF document
Available under license: CC BY-NC-ND: Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License
Final published version
Licence: CC BY-NC-ND: Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License
Research output: Contribution to Journal/Magazine › Journal article › peer-review
Research output: Contribution to Journal/Magazine › Journal article › peer-review
}
TY - JOUR
T1 - A2-FPN for semantic segmentation of fine-resolution remotely sensed images
AU - Li, Rui
AU - Wang, Libo
AU - Zhang, Ce
AU - Duan, Chenxi
AU - Zheng, Shunyi
PY - 2022/2/28
Y1 - 2022/2/28
N2 - The thriving development of earth observation technology makes more and more high-resolution remote-sensing images easy to obtain. However, caused by fine-resolution, the huge spatial and spectral complexity leads to the automation of semantic segmentation becoming a challenging task. Addressing such an issue represents an exciting research field, which paves the way for scene-level landscape pattern analysis and decision-making. To tackle this problem, we propose an approach for automatic land segmentation based on the Feature Pyramid Network (FPN). As a classic architecture, FPN can build a feature pyramid with high-level semantics throughout. However, intrinsic defects in feature extraction and fusion hinder FPN from further aggregating more discriminative features. Hence, we propose an Attention Aggregation Module (AAM) to enhance multiscale feature learning through attention-guided feature aggregation. Based on FPN and AAM, a novel framework named Attention Aggregation Feature Pyramid Network (A2-FPN) is developed for semantic segmentation of fine-resolution remotely sensed images. Extensive experiments conducted on four datasets demonstrate the effectiveness of our A2-FPN in segmentation accuracy. Code is available at https://github.com/lironui/A2-FPN.
AB - The thriving development of earth observation technology makes more and more high-resolution remote-sensing images easy to obtain. However, caused by fine-resolution, the huge spatial and spectral complexity leads to the automation of semantic segmentation becoming a challenging task. Addressing such an issue represents an exciting research field, which paves the way for scene-level landscape pattern analysis and decision-making. To tackle this problem, we propose an approach for automatic land segmentation based on the Feature Pyramid Network (FPN). As a classic architecture, FPN can build a feature pyramid with high-level semantics throughout. However, intrinsic defects in feature extraction and fusion hinder FPN from further aggregating more discriminative features. Hence, we propose an Attention Aggregation Module (AAM) to enhance multiscale feature learning through attention-guided feature aggregation. Based on FPN and AAM, a novel framework named Attention Aggregation Feature Pyramid Network (A2-FPN) is developed for semantic segmentation of fine-resolution remotely sensed images. Extensive experiments conducted on four datasets demonstrate the effectiveness of our A2-FPN in segmentation accuracy. Code is available at https://github.com/lironui/A2-FPN.
KW - semantic segmentation
KW - deep learning
KW - attention mechanism
U2 - 10.1080/01431161.2022.2030071
DO - 10.1080/01431161.2022.2030071
M3 - Journal article
VL - 43
SP - 1131
EP - 1155
JO - International Journal of Remote Sensing
JF - International Journal of Remote Sensing
SN - 0143-1161
IS - 3
ER -