Learnable Quantum Efficiency Filters for Urban Hyperspectral Segmentation

2026-03-27

AI Summary

This work addresses the challenges of poor interpretability and low learning efficiency in semantic segmentation of high-dimensional urban hyperspectral data. The authors propose a Learnable Quantum Efficiency (LQE) method that, for the first time, integrates physical constraints—such as single dominant peak, smoothness, and bandwidth limitations—into an end-to-end trainable spectral response function. This approach achieves parameter-efficient and interpretable hyperspectral dimensionality reduction, requiring only 12–36 parameters, substantially fewer than existing learnable methods (51–22K). Evaluated on three urban hyperspectral datasets, LQE improves mean Intersection-over-Union (mIoU) by up to 2.45% over conventional approaches and by up to 1.56% over current learnable methods, while maintaining controllable inference latency. The method establishes a new paradigm for hyperspectral perception and multispectral sensor design.

Abstract

Hyperspectral sensing provides rich spectral information for scene understanding in urban driving, but its high dimensionality poses challenges for interpretation and efficient learning. We introduce Learnable Quantum Efficiency (LQE), a physics-inspired, interpretable dimensionality reduction (DR) method that parameterizes smooth high-order spectral response functions that emulate plausible sensor quantum efficiency curves. Unlike conventional methods or unconstrained learnable layers, LQE enforces physically motivated constraints, including a single dominant peak, smooth responses, and bounded bandwidth. This formulation yields a compact spectral representation that preserves discriminative information while remaining fully differentiable and end-to-end trainable within semantic segmentation models (SSMs). We conduct systematic evaluations across three publicly available multi-class hyperspectral urban driving datasets, comparing LQE against six conventional and seven learnable baseline DR methods across six SSMs. Averaged across all SSMs and configurations, LQE achieves the highest average mIoU, improving over conventional methods by 2.45\%, 0.45\%, and 1.04\%, and over learnable methods by 1.18\%, 1.56\%, and 0.81\% on HyKo, HSI-Drive, and Hyperspectral City, respectively. LQE maintains strong parameter efficiency (12--36 parameters compared to 51--22K for competing learnable approaches) and competitive inference latency. Ablation studies show that low-order configurations are optimal, while the learned spectral filters converge to dataset-intrinsic wavelength patterns. These results demonstrate that physics-informed spectral learning can improve both performance and interpretability, providing a principled bridge between hyperspectral perception and data-driven multispectral sensor design for automotive vision systems.

Problem

• hyperspectral segmentation
• dimensionality reduction
• urban driving
• spectral representation
• quantum efficiency

Innovation

• Learnable Quantum Efficiency
• Physics-informed Spectral Learning
• Hyperspectral Segmentation
• Interpretable Dimensionality Reduction
• Sensor-inspired Neural Design