How Does Urban Context Relate to Residential Building Health? A Vision-POI Fusion Framework for Building-Level Housing Inspection

2026-07-23 · Source: cs.CV updates on arXiv.org · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics, Geographic Information Systems · Depth: Advanced, extended

Summary

A vision–point-of-interest (POI) fusion framework has been developed for residential building health assessment, combining multi-view visual inspection with POI-derived neighborhood context. This framework was empirically validated using a dataset from Qingdao, China, encompassing 92 old residential communities, 3,237 buildings, and 25,608 field-acquired images across seven housing issue categories. The methodology involves evaluating multiple object detection models to extract image-level issue data, which is then aggregated into interpretable building-level representations like detection frequency and confidence statistics. Concurrently, POI features are extracted within 500m, 1000m, and 1500m neighborhood buffers to characterize surrounding functional environments. These visual and POI features are integrated using a cost-sensitive Random Forest classifier. Results show multi-view visual aggregation significantly improved building-level Macro-F1 from 60.84% to 74.95%, with POI context providing an additional, modest gain to 76.79%, functioning as a supplementary prior.

Key takeaway

For urban planners or property managers implementing scalable housing inspection programs, you should prioritize multi-view visual aggregation for primary detection accuracy. Supplement this with 1000m POI data to refine building-level judgments, especially for ambiguous cases. This approach improves detection robustness and provides spatially interpretable insights for targeted urban renewal governance. Be aware that POI gains are modest and category-dependent, not replacing direct visual evidence.

Key insights

Fusing multi-view visual data with urban POI context enhances building health assessment accuracy.

Principles

• Multi-view aggregation significantly improves building-level inspection accuracy.
• Urban context from POIs offers supplementary, not causal, predictive value.
• Optimal POI buffer radius balances coverage and spatial specificity.

Method

A three-stage framework: visual perception (object detection, multi-view aggregation), spatial association (POI feature extraction, correlation analysis), and contextual post-correction (Random Forest classifier fusion).

In practice

• Aggregate image detections to building-level features for robust assessment.
• Use 1000m POI buffers for optimal contextual feature integration.
• Employ cost-sensitive Random Forest for imbalanced issue categories.

Topics

• Residential Building Health
• Urban Physical Examination
• Object Detection
• Multi-View Aggregation
• Points of Interest
• Visual-GIS Fusion

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

Related on AIssential

• Life Style Levels: Neighborhood Delineation using Geospatial Data — Open-source satellite imagery and building morphology enable scalable, fine-scale urban socioeconomic mapping in developing regions.
• Context-Aware Feature-Fusion for Co-occurring Object Detection in Autonomous Driving — Context-aware feature fusion using local and global attention significantly improves object detection for challenging cases in autonomous driving.
• ReA-OVCD: Reliability-Aware Open-Vocabulary Change Detection via Semantic and Spatial Refinement — ReA-OVCD refines open-vocabulary change detection by collaboratively addressing semantic ambiguity and spatial inconsistency through a training-free, two-module approach.
• A Study of Failure Modes in Two-Stage Human-Object Interaction Detection — Two-stage HOI models struggle with complex scenes and rare interactions despite high benchmark scores.
• Toward Seasonal Guidelines for Robust Deep-Learning Sentinel-2 Building Detection in Different Area Types — Seasonal and built-up environment variations significantly impact deep learning model performance for Sentinel-2 building detection.
• Global Building Area Estimation Products: How Accurate Are They? — Independent evaluation reveals global building area product accuracy varies significantly by region and methodology.

See Counsel's argued verdicts on the open AI decisions leaders are weighing →

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.CV updates on arXiv.org.