Shape from Polarization of Thermal Emission and Reflection

2025-06-22

AI Summary

Estimating the 3D shape of transparent objects remains challenging due to complex light transport, and existing long-wave infrared (LWIR) polarization-based shape-from-polarization (SfP) methods suffer from significant errors—primarily because they neglect thermal reflection effects and employ incomplete polarization modeling. To address this, we propose the first physically grounded LWIR polarization imaging model that explicitly jointly models both thermal emission and surface reflection. We further introduce ThermoPol, the first real-world LWIR SfP benchmark dataset, and design a physics-guided synthetic data generation strategy to train a hybrid model-driven and learning-based neural network for high-accuracy surface normal estimation. Experiments demonstrate that our method substantially reduces the systematic errors induced by reflection omission in conventional approaches across diverse transparent materials, exhibits strong generalization and practical applicability, and establishes a new paradigm for LWIR polarization vision.

Technology Category

• Computer Vision: Low Level & Physics-based Vision
• Machine Learning: Large Multimodal Models (LMMs)
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Security and Privacy: Data transparency and provenance
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Shape estimation for transparent objects is challenging due to their complex light transport. To circumvent these difficulties, we leverage the Shape from Polarization (SfP) technique in the Long-Wave Infrared (LWIR) spectrum, where most materials are opaque and emissive. While a few prior studies have explored LWIR SfP, these attempts suffered from significant errors due to inadequate polarimetric modeling, particularly the neglect of reflection. Addressing this gap, we formulated a polarization model that explicitly accounts for the combined effects of emission and reflection. Based on this model, we estimated surface normals using not only a direct model-based method but also a learning-based approach employing a neural network trained on a physically-grounded synthetic dataset. Furthermore, we modeled the LWIR polarimetric imaging process, accounting for inherent systematic errors to ensure accurate polarimetry. We implemented a prototype system and created ThermoPol, the first real-world benchmark dataset for LWIR SfP. Through comprehensive experiments, we demonstrated the high accuracy and broad applicability of our method across various materials, including those transparent in the visible spectrum.

Problem

• Modeling polarization for transparent objects in LWIR
• Addressing errors from inadequate polarimetric modeling
• Creating accurate surface normals via emission and reflection

Innovation

• LWIR SfP with emission and reflection modeling
• Neural network trained on synthetic dataset
• ThermoPol benchmark dataset for LWIR SfP