Reproduce “Building Height Estimation Using Shadow Length in Satellite Imagery” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Mahd Qureshi, Shayaan Chaudhry, Sana Jabba, Murtaza Taj
This paper estimates building heights from single monocular satellite images by analyzing the shadows cast by structures. The method integrates deep learning models for building detection and shadow length regression with a mathematical formula linking shadow length to height and solar elevation angle, achieving an RMSE of 3.84 meters on a dataset covering 42 Chinese cities.
Key insights
- Building height can be accurately estimated from monocular satellite imagery by leveraging shadow length, which is directly related to height and solar elevation angle.
- Integrating deep learning for object detection and shadow length regression with a physical mathematical model provides both performance improvements and interpretability.
- A custom annotation tool and a shadow-based height loss function are crucial for training models on shadow-derived height data, particularly when image metadata is incomplete.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.