End-to-End Optimization of Polarimetric Measurement and Material Classifier

2026-03-20

AI Summary

Traditional polarimetric measurements rely on multiple modulations of the incident light’s polarization state, resulting in a lengthy process that often includes redundant information for classification tasks, with no clear consensus on optimal measurement configurations. This work proposes the first end-to-end joint optimization framework that simultaneously learns a material classifier and an optimal set of rotation angles for polarization optics, automatically identifying the minimal subset of effective polarimetric measurements tailored to the classification objective. The approach builds upon a differentiable polarimetric optical model based on the Mueller matrix formalism and integrates deep learning to jointly optimize measurement strategy and classification performance. Experimental results demonstrate that the method significantly reduces the number of required measurements while maintaining high classification accuracy, thereby substantially improving the efficiency of polarimetric sensing.

Technology Category

• Computer Vision: Learning & Optimization for CV
• Machine Learning: Optimization
• Search and Optimization: Learning to Search

Application Category

• Security and Privacy: Large-scale security measurements
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Material classification is a fundamental problem in computer vision and plays a crucial role in scene understanding. Previous studies have explored various material recognition methods based on reflection properties such as color, texture, specularity, and scattering. Among these cues, polarization is particularly valuable because it provides rich material information and enables recognition even at distances where capturing high-resolution texture is impractical. However, measuring polarimetric reflectance properties typically requires multiple modulations of the polarization state of the incident light, making the process time-consuming and often unnecessary for certain recognition tasks. While material classification can be achieved using only a subset of polarimetric measurements, the optimal configuration of measurement angles remains unclear. In this study, we propose an end-to-end optimization framework that jointly learns a material classifier and determines the optimal combinations of rotation angles for polarization elements that control both the incident and reflected light states. Using our Mueller-matrix material dataset, we demonstrate that our method achieves high-accuracy material classification even with a limited number of measurements.

Problem

• material classification
• polarimetric measurement
• optimal configuration
• polarization angles
• Mueller matrix

Innovation

• end-to-end optimization
• polarimetric measurement
• material classification
• Mueller matrix
• polarization angles