The rapid evolution of optical metrology-including camera-based systems, confocal sensing, interferometry, and tomography—has significantly improved the precision, flexibility, and applicability of modern measurement systems. These technologies are widely used in advanced manufacturing, biomedical imaging, and scientific research, enabling high-resolution, multidimensional, non-contact measurements, and establishing optical metrology as a key tool for industrial inspection and experimental analysis. However, real-world measurement scenarios introduce substantial challenges. Optical data are affected by multiple noise sources, such as photon shot noise, electronic interference, environmental disturbances, and optical aberrations, which degrade signal quality. Measurements also often include missing or corrupted data due to occlusions, limited viewpoints, surface reflectivity, or hardware limitations. These issues are exacerbated by the need for real-time processing and low-latency decision-making, especially in industrial and clinical settings handling large data volumes.

Traditional signal processing and calibration methods, while well established, struggle to cope with this variability and scale. They require extensive manual tuning, are sensitive to changes in operating conditions, and often generalize poorly across setups. Meanwhile, modern systems demand not only efficiency but also interpretability, reliability, and metrological traceability aligned with physical principles and uncertainty constraints. Recent advances in machine learning and deep learning offer promising solutions, showing strong performance in denoising, reconstruction, defect detection, and analysis of high-dimensional data. Nonetheless, challenges remain, including limited labeled data, robustness across instruments, model interpretability, and validation against established physical standards.

Physics-Informed AI, Measurement Uncertainty, Metrological Traceability, Data Reconstruction, Explainable Artificial Intelligence, Precision Metrology