TY - JOUR
T1 - Deep Learning for the Automatic Division of Building Constructions into Sections on Remote Sensing Images
AU - Schuegraf, Philipp
AU - Zorzi, Stefano
AU - Fraundorfer, Friedrich
AU - Bittner, Ksenia
N1 - Publisher Copyright:
© 2008-2012 IEEE.
PY - 2023
Y1 - 2023
N2 - Urban areas predominantly consist of complex building structures, which are assembled of multiple building sections. From very high resolution remote sensing imagery, not only roof-tops but also the separation lines between them are visible. Since fully convolutional neural network (FCN)-based methods have become the primary choice in segmentation approaches, they have been extensively used for automatic building footprint extraction. But each of the previous works on building segmentation either lacks separation of building blocks into sections or does not produce sections of regular appearance. We propose a two-stage approach to overcome these limitations. The first step segments building and separation lines using an FCN model and the second step produces building instances by using a learning-free method. Our model receives a top-down image and a digital surface model (DSM) patch in two separate encoders. The encoder features are summed before the skip connections, which utilize the encoder features from the current and higher-resolution feature maps. We train our model with regularization losses for building shapes and separation lines on both satellite and aerial imagery. We test our model on a city that was not previously included in the training phase to show that it has the capacity to generalize across different geographical locations and architectural styles. Furthermore, we use our building section instance predictions to generate: 1) vectorized building maps and 2) a level-of-detail-1 DSM.
AB - Urban areas predominantly consist of complex building structures, which are assembled of multiple building sections. From very high resolution remote sensing imagery, not only roof-tops but also the separation lines between them are visible. Since fully convolutional neural network (FCN)-based methods have become the primary choice in segmentation approaches, they have been extensively used for automatic building footprint extraction. But each of the previous works on building segmentation either lacks separation of building blocks into sections or does not produce sections of regular appearance. We propose a two-stage approach to overcome these limitations. The first step segments building and separation lines using an FCN model and the second step produces building instances by using a learning-free method. Our model receives a top-down image and a digital surface model (DSM) patch in two separate encoders. The encoder features are summed before the skip connections, which utilize the encoder features from the current and higher-resolution feature maps. We train our model with regularization losses for building shapes and separation lines on both satellite and aerial imagery. We test our model on a city that was not previously included in the training phase to show that it has the capacity to generalize across different geographical locations and architectural styles. Furthermore, we use our building section instance predictions to generate: 1) vectorized building maps and 2) a level-of-detail-1 DSM.
KW - Convolutional neural networks
KW - deep learning
KW - semantic segmentation
KW - supervised learning
KW - urban areas
UR - http://www.scopus.com/inward/record.url?scp=85165302287&partnerID=8YFLogxK
U2 - 10.1109/JSTARS.2023.3296449
DO - 10.1109/JSTARS.2023.3296449
M3 - Article
AN - SCOPUS:85165302287
SN - 1939-1404
VL - 16
SP - 7186
EP - 7200
JO - IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
JF - IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
ER -