Arbitrary-Shaped Building Boundary-Aware Detection with Pixel Aggregation Network

Jiang，Xin1; Zhang，Xinchang2; Xin，Qinchuan3; Xi，Xu4; Zhang，Pengcheng5

2020

10.1109/JSTARS.2020.3017934

IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing   影响因子和分区

1939-1404

2151-1535

Large-scale building extraction is an essential work in the field of remote sensing image analysis. The high-resolution image extraction methods based on deep learning have achieved state-of-the-art performance. However, most of the previous work has focused on region accuracy rather than boundary quality. Aiming at the low accuracy problems and incomplete boundary of the building extraction method, we propose an architecture that optimizes the boundaries of buildings, BAPANet. Notably, the architecture consists of an encoder-decoder network and residual refinement modules responsible for prediction and refinement, respectively. The objective function optimizes the network in the form of three levels (pixel, feature map, and patch) by fusing three loss functions: binary cross-entropy (BCE), intersection over-union (IoU) and structural similarity (SSIM). The five public datasets' experimental results show that the extraction method in this paper has high region accuracy, and the boundary of buildings is clear and complete.

boundary quality Building extraction Buildings Feature extraction high-resolution Image edge detection Image segmentation Optimization Remote sensing Semantics structural similarity

英语

第一

2-s2.0-85102640774

Scopus

1.School of Environmental Science & Engineering, Southern University of Science and Technology, 255310 Shenzhen, Guangdong China (e-mail: jiangxiaobaix@outlook.com)
2.School of Geographical Sciences, Guangzhou University, 47875 Guangzhou, Guangdong China (e-mail: eeszxc@mail.sysu.edu.cn)
3.Department of Geography and Planning, Sun Yat-Sen University, 26469 Guangzhou, Guangdong China (e-mail: xinqinchuan@mail.sysu.edu.cn)
4.School of Environmental Science and Engineering, Suzhou University of Science and Technology, 66339 Suzhou, Jiangsu China (e-mail: xixu2016sysu@outlook.com)
5.Guangzhou Urban planing & design survey research institute, Guangzhou University, 47875 Guangzhou, Guangdong China (e-mail: 232323@qq.com)

Jiang，Xin,Zhang，Xinchang,Xin，Qinchuan,et al. Arbitrary-Shaped Building Boundary-Aware Detection with Pixel Aggregation Network[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing,2020.

Jiang，Xin,Zhang，Xinchang,Xin，Qinchuan,Xi，Xu,&Zhang，Pengcheng.(2020).Arbitrary-Shaped Building Boundary-Aware Detection with Pixel Aggregation Network.IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing.

Jiang，Xin,et al."Arbitrary-Shaped Building Boundary-Aware Detection with Pixel Aggregation Network".IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing (2020).