Quantifying Potential Risk through Passively Sensed Invisible Modalities

Risk manifests not only through visible changes but also through invisible signals such as overheating, abnormal vibrations, and irregular noise, necessitating proactive risk management. Despite advances in neural network-driven visual risk assessment, overreliance on visual modality limits comprehensive risk detection, as early-stage risks often lack clear visual indicators. This paper proposes a risk quantification framework leveraging passive sensing of invisible modalities, including thermal, vibration and acoustic signals, to enhance early risk detection. Passive sensing minimizes environmental interference while capturing latent risk indicators. The framework’s effectiveness is validated through three case studies: quality risk quantification via thermal characteristics in PCBA inspection, fatigue risk quantification via vibration data for machinery condition assessment, and collision risk quantification via acoustic signals for NLoS vehicle localization. This approach advances risk assessment by integrating passively sensed invisible modalities beyond traditional visual-based methods.