Generative building feature estimation from satellite images
IEEE Transactions on Geoscience and Remote Sensing

Abstract

Urban and environmental researchers seek to obtain building features (e.g., building shapes, counts, and areas) at large scales. However, blurriness, occlusions, and noise from prevailing satellite images severely hinder the performance of image segmentation, super-resolution, or deep-learning-based translation networks. In this article, we combine globally available satellite images and spatial geometric feature datasets to create a generative modeling framework that enables obtaining significantly improved accuracy in per-building feature estimation and the generation of visually plausible building footprints. Our approach is a novel design that compensates for the degradation present in satellite images by using a novel deep network setup that includes segmentation, generative modeling, and adversarial learning for instance-level building features. Our method has proven its robustness through large-scale prototypical experiments covering heterogeneous scenarios from dense urban to sparse rural. Results show better quality over advanced segmentation networks for urban and environmental planning, and show promise for future continental-scale urban applications.

Method

We use a Adversarial Variational Autoencoder to generate building layout from Low-Resolution but public accessible satellite imagery. The method produces good instance-level behavior of building footprint and features despite extreme blurriness and occlusions from satellite imagery.

Comparisons to SOTA

Compare to Microsoft Building Footprints

Large-Scale Implementation in Belgium

BibTeX

@article{he2023generative,

    title={Generative building feature estimation from satellite images},

    author={He, Liu and Shan, Jie and Aliaga, Daniel},

    journal={IEEE Transactions on Geoscience and Remote Sensing}, 

    volume={61}, 

    pages={1--13}, 

    year={2023},

    publisher={IEEE}

  }

Acknowledgements

Thanks Dr. Jacques Teller for providing Belgium dataset.

The website template was borrowed from Michaël Gharbi.