An Object-Based Deep Learning Approach for Building Height Estimation from Single SAR Images

2025-07-10

AI Summary

To address the challenge of estimating building heights from single high-resolution SAR imagery, this paper proposes an object-oriented deep learning regression framework: first detecting building bounding boxes, then performing pixel-wise height regression within each detected region. The method jointly integrates object detection and regression estimation, trained end-to-end on COSMO-SkyMed SAR data. To enhance generalizability, a cross-city cross-validation strategy is introduced during training. Its key innovation lies in establishing an object-level modeling paradigm, which significantly improves model transferability and robustness to out-of-distribution (cross-continental) urban scenes. Evaluated across multiple European cities, the method achieves a mean absolute error (MAE) of 2.20 m—surpassing state-of-the-art approaches—and demonstrates exceptional reliability even in complex urban environments with dense, heterogeneous structures.

Technology Category

• Computer Vision: Object Detection & Categorization
• Intelligent Robots: State Estimation
• Machine Learning: Classification and Regression

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Systems and Infrastructure for Web, Mobile and WoT: Experiences and lessons learnt from Web-based algorithms and system deployments
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs

Abstract

Accurate estimation of building heights using very high resolution (VHR) synthetic aperture radar (SAR) imagery is crucial for various urban applications. This paper introduces a Deep Learning (DL)-based methodology for automated building height estimation from single VHR COSMO-SkyMed images: an object-based regression approach based on bounding box detection followed by height estimation. This model was trained and evaluated on a unique multi-continental dataset comprising eight geographically diverse cities across Europe, North and South America, and Asia, employing a cross-validation strategy to explicitly assess out-of-distribution (OOD) generalization. The results demonstrate highly promising performance, particularly on European cities where the model achieves a Mean Absolute Error (MAE) of approximately one building story (2.20 m in Munich), significantly outperforming recent state-of-the-art methods in similar OOD scenarios. Despite the increased variability observed when generalizing to cities in other continents, particularly in Asia with its distinct urban typologies and prevalence of high-rise structures, this study underscores the significant potential of DL for robust cross-city and cross-continental transfer learning in building height estimation from single VHR SAR data.

Problem

• Estimating building heights from single SAR images
• Improving cross-continental generalization in height estimation
• Reducing errors in urban building height prediction

Innovation

• Object-based deep learning for SAR height estimation
• Bounding box detection followed by regression
• Cross-continental dataset for OOD generalization