Indoor Airflow Imaging Using Physics-Informed Background-Oriented Schlieren Tomography

2025-09-17

AI Summary

This work addresses the challenging problem of single-view, non-intrusive reconstruction of indoor 3D airflow fields. We propose a novel method integrating Background-Oriented Schlieren (BOS) imaging with Physics-Informed Neural Networks (PINNs). By modeling optical distortions of projected patterns induced by refractive index variations in inhomogeneous airflow, our approach jointly reconstructs the 3D refractive index field and enforces physical consistency via buoyancy-driven Navier–Stokes equations. We design a differentiable ray-tracing and physics-driven rendering pipeline, incorporating PINN-based regularization derived from governing PDEs and a multi-physical loss function coupling velocity, density, and optical measurements. Our method overcomes the ill-posedness inherent in single-view tomography, enabling high-fidelity, calibration-free, tracer-free 3D reconstruction of velocity and density fields. Experiments demonstrate accurate recovery of complex indoor flow features—including thermal plumes and vortical structures—achieving significant improvements in reconstruction accuracy and physical fidelity.

Technology Category

• Computer Vision: Low Level & Physics-based Vision
• Planning, Routing, and Scheduling: Activity and Plan Recognition
• Search and Optimization: Learning to Search

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs

Abstract

We develop a framework for non-invasive volumetric indoor airflow estimation from a single viewpoint using background-oriented schlieren (BOS) measurements and physics-informed reconstruction. Our framework utilizes a light projector that projects a pattern onto a target back-wall and a camera that observes small distortions in the light pattern. While the single-view BOS tomography problem is severely ill-posed, our proposed framework addresses this using: (1) improved ray tracing, (2) a physics-based light rendering approach and loss formulation, and (3) a physics-based regularization using a physics-informed neural network (PINN) to ensure that the reconstructed airflow is consistent with the governing equations for buoyancy-driven flows.

Problem

• Non-invasive volumetric indoor airflow estimation
• Single-view background-oriented schlieren tomography
• Physics-informed reconstruction of buoyancy-driven flows

Innovation

• Physics-informed neural network regularization
• Improved ray tracing technique
• Physics-based light rendering approach