The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences
Download
Publications Copernicus
Download
Citation
Articles | Volume XLIII-B1-2022
Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., XLIII-B1-2022, 39–44, 2022
https://doi.org/10.5194/isprs-archives-XLIII-B1-2022-39-2022
Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., XLIII-B1-2022, 39–44, 2022
https://doi.org/10.5194/isprs-archives-XLIII-B1-2022-39-2022
 
30 May 2022
30 May 2022

A SHORT-CUT CONNECTIONS-BASED NEURAL NETWORK FOR BUILDING EXTRACTION FROM HIGH RESOLUTION ORTHOIMAGERY

Z. He1, H. He2, J. Li2, M. A. Chapman3, and H. Ding1 Z. He et al.
  • 1Nanjing University of Information Science and Technology, Nanjing, 210044, China
  • 2Department of Geography and Environmental Management, University of Waterloo, Waterloo, ON N2L 3G1, Canada
  • 3Department of Civil Engineering, Geomatics Engineering, Ryerson University, Toronto, NO M5B 2K3, Canada

Keywords: Deep Learning, Building Extraction, Dilated Convolution, Short-cut Connections, High Resolution Orthoimagery

Abstract. Extracting building footprints utilizing deep learning-based (DL-based) methods for high-resolution remote sensing images is one of the current research interest areas. However, the extraction results suffer from blurred edges, rounded corners and detail loss in general. Hence, this article presents a detail-oriented deep learning network named eU-Net (enhanced U-Net). The method adopted in this study, imagery send into the pre-module, which consists of the Canny edge detector, Principal Component Analysis (PCA) and the inter-band ratio operations, before feeding them into the network. Then, process skips connections used in the network to reduce the loss of details during edge and corner detection. The encoding and decoding modules, in this network, are redesigned to expand the perceptual field with shortcut connections and stacked layers. Finally, a Dropout module is added in the bottom layer of the network to avoid the over-fitting problem. The experimental results indicate that the methods used in this study outperform other commonly used and state-of-the-art methods of FCN-8s, U-net, DeepLabv3 and Fast SCNN.