Multi-scale building instance refinement extraction from remote sensing images by fusing with decentralized adaptive attention mechanism

被引：0
作者：
Jiang B. [1 ,2 ]
Hang W. [2 ]
Xu S. [2 ]
Wu Y. [2 ]
机构：
[1] School of Computer Science, China University of Geosciences, Wuhan
[2] National Engineering Research Center of Geographic Information System, China University of Geosciences, Wuhan
来源：
Cehui Xuebao/Acta Geodaetica et Cartographica Sinica | 2023年 / 52卷 / 09期
基金：
中国国家自然科学基金;
关键词：
adaptive attention mechanism; building refinement extraction; deep learning; multi-scale; remote sensing images; split-attention networks;
D O I：
10.11947/j.AGCS.2023.20220322
中图分类号：
学科分类号：
摘要：
The accurate and efficient automatic extraction of building footprints from remote sensing images has a wide range of applications. Since the buildings in remote sensing images have different types, scales, shapes and backgrounds, the existing methods, to varying degrees, suffer from the problems of missing small-scale buildings, blurred contour boundaries, and inability to distinguish individual building instances. Therefore, this paper proposed a multi-scale building instance refinement extraction convolutional neural network (MBRef-CNN) fusing with decentralized adaptive attention mechanism for remote sensing images. First, a feature pyramid network fused with split-attention and adaptive attention mechanism (SA-FPN) was used to learn multi-scale building features. Then, according to the multi-scale features, the region proposal network (RPN) was used to detect the location of individual building instances. Finally, the boundary refinement network (BndRN) was used to iteratively acquire the precise building masks. On WHU aerial imagery dataset, the comparison experiments were conducted with the existing popular segmentation methods. The results show that the accuracy of the proposed method in this paper is higher than the others. Moreover, the MBRef-CNN shows good comprehensive performance in multi-scale building extraction, and has obvious accuracy advantages in small-scale building extraction. © 2023 SinoMaps Press. All rights reserved.
引用
收藏
页码：1504 / 1514
页数：10
相关论文
共 28 条
[11]  
HU Jie, SHEN Li, ALBANIE S, Et al., Squeeze-and-excitation networks, IEEE Transactions on Pattern Analysis and Machine Intelligence, 42, 8, pp. 2011-2023, (2020)
[12]  
ZHU Qing, LIAO Cheng, HU Han, Et al., MAP-net: multiple attending path neural network for building footprint extraction from remote sensed imagery, IEEE Transactions on Geoscience and Remote Sensing, 59, 7, pp. 6169-6181, (2021)
[13]  
HE Kaiming, GKIOXARI G, DOLLAR P, Et al., Mask R-CNN, Proceedings of 2017 IEEE International Conference on Computer Vision, pp. 2980-2988, (2017)
[14]  
REN Shaoqing, HE Kaiming, GIRSHICK R, Et al., Faster R-CNN: towards real-time object detection with region proposal networks, IEEE Transactions on Pattern Analysis and Machine Intelligence, 39, 6, pp. 1137-1149, (2017)
[15]  
CHEN Kai, PANG Jiangmiao, WANG Jiaqi, Et al., Hybrid task cascade for instance segmentation, Proceedings of 2019 IEEE Conference on Computer Vision and Pattern Recognition, pp. 4969-4978, (2019)
[16]  
GAO Naiyu, SHAN Yanhu, WANG Yupei, Et al., SSAP: single-shot instance segmentation with affinity pyramid, Proceedings of 2019 IEEE International Conference on Computer Vision, pp. 642-651, (2019)
[17]  
ARNAB A, TORR P H S., Pixelwise instance segmentation with a dynamically instantiated network, Proceedings of 2017 IEEE Conference on Computer Vision and Pattern Recognition, pp. 441-450, (2017)
[18]  
SHRESTHA S, VANNESCHI L., Improved fully convolutional network with conditional random fields for building extraction, Remote Sensing, 10, 7, (2018)
[19]  
YANG Guang, ZHANG Qian, ZHANG Guixu, EANet: edge-aware network for the extraction of buildings from aerial images, Remote Sensing, 12, 13, (2020)
[20]  
JUNG H, CHOI H S, KANG M., Boundary enhancement semantic segmentation for building extraction from remote sensed image, IEEE Transactions on Geoscience and Remote Sensing, 60, pp. 1-12, (2022)
123