Building Extraction from Remote Sensing Imagery under Hazy and Low-light Conditions: Benchmark and Baseline

2026-04-16 · Source: Computer Vision and Pattern Recognition · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Remote Sensing · Depth: Expert, quick

Summary

HaLoBuilding is introduced as the first optical benchmark dataset specifically designed for building extraction from remote sensing (RS) imagery under challenging hazy and low-light conditions. Existing methods and benchmarks primarily focus on ideal clear-weather scenarios, leading to performance degradation in real-world applications. This new benchmark utilizes a same-scene multitemporal pairing strategy to ensure pixel-level label alignment and high fidelity, even under extreme atmospheric degradation. Alongside the dataset, the authors propose HaLoBuild-Net, an end-to-end framework for building extraction in adverse RS environments. HaLoBuild-Net incorporates a Spatial-Frequency Focus Module (SFFM) to mitigate meteorological interference, a Global Multi-scale Guidance Module (GMGM) for global semantic constraints, and a Mutual-Guided Fusion Module (MGFM) for semantic-spatial calibration. Experiments show HaLoBuild-Net significantly outperforms other methods on HaLoBuilding and generalizes well to WHU, INRIA, and LoveDA datasets.

Key takeaway

For computer vision engineers developing building extraction models, the HaLoBuilding dataset and HaLoBuild-Net offer a critical advancement for real-world applications. Your current models likely underperform in hazy or low-light conditions; adopting this benchmark and framework can significantly improve robustness and accuracy in adverse weather scenarios. Consider integrating the proposed modules to enhance feature extraction and semantic consistency in your next-generation models.

Key insights

A new benchmark and network address building extraction challenges in hazy and low-light remote sensing imagery.

Principles

• Multitemporal pairing ensures pixel-level label alignment.
• Coupling spatial and frequency domains mitigates interference.

Method

HaLoBuild-Net uses SFFM for meteorological interference mitigation, GMGM for global semantic constraints, and MGFM for bidirectional semantic-spatial calibration to extract buildings from degraded RS imagery.

In practice

• Utilize HaLoBuilding dataset for robust model training.
• Integrate SFFM for weather-robust feature extraction.
• Apply GMGM for improved topological consistency.

Topics

• Building Extraction
• Remote Sensing
• Hazy & Low-light Conditions
• HaLoBuilding Dataset
• HaLoBuild-Net

Code references

• AeroVILab-AHU/HaLoBuilding

Best for: AI Scientist, Computer Vision Engineer, Research Scientist

Related on AIssential

• PCFootprint: A Large-Scale Dataset and Benchmark for Vectorized Building Footprint Extraction from Aerial LiDAR Point Clouds — PCFootprint is the first large-scale LiDAR dataset for vectorized building footprint extraction, addressing optical imagery limitations.
• VistaHop: Benchmarking Multi-hop Visual Reasoning for Visual DeepSearch — VistaHop benchmarks multi-hop visual reasoning, revealing current MLRM limitations in complex visual search tasks.
• AIGS-Net: Compact Illumination Field Modeling via 2D Gaussian Splatting for Fast Low-Light Image Enhancement — AIGS-Net uses adaptive 2D Gaussian Splatting and zero-parameter encoding for ultra-lightweight, fast low-light image enhancement.
• Beyond Visual Fidelity: Benchmarking Super-Resolution Models for Large-Scale Remote Sensing Imagery via Downstream Task Integration — Traditional SR metrics often fail to predict real-world performance in downstream remote sensing tasks.
• Semantic Smoothing via Novel View Synthesis for Robust SAR Image Classification — Semantic smoothing uses generative models to create semantically consistent views for robust classification, outperforming isotropic noise.
• Monocular Building Height Estimation from PhiSat-2 Imagery: Dataset and Method — PhiSat-2 imagery, combined with a novel two-stream ordinal network, significantly improves monocular building height estimation.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Computer Vision and Pattern Recognition.