Development of an infrared camera system with an integrated polarizing filter for omnidirectional thermal background reflection removal

• A rotating wire-grid polarizer enables omnidirectional thermal reflection suppression. • Pixel-wise sinusoidal modeling separates reflection and radiative components. • No need for selecting a reflection-free polarization angle. • Significant reduction of false positives in daytime bridge inspections. • Comparable NETD performance achieved with an uncooled infrared camera. Thermal background reflection remains a major obstacle to reliable daytime infrared thermography (IRT) for infrastructure inspection. Specular and polarization-dependent reflections often generate pseudo temperature anomalies, leading to false-positive detections and reduced interpretability. This study presents the development of a practical infrared camera system integrating a uniformly rotating wire-grid polarizing filter for omnidirectional thermal background reflection removal. The filter rotates at 2 Hz while thermal images are acquired at 16 frames per second. Sixteen consecutive frames acquired over two full rotations (1 s) are used as one analysis cycle, enabling uniform polarization sampling and stable temporal modeling. For each pixel, the temperature variation within this 16-frame sequence is modeled using sinusoidal fitting, allowing separation of the polarization-dependent reflection component (amplitude) from the baseline radiative temperature (offset). The reflection-suppressed temperature is derived by removing the periodic component without relying on identification of a specific reflection-free polarization angle. Laboratory validation using multi-material specimens (concrete, PVC repair sheet, free lime, and coating) demonstrated material-dependent reflection behavior and effective artifact suppression. Field validation on an in-service bridge confirmed stable daytime operation and significant reduction of reflection-induced false positives, while preserving all previously confirmed critical regions. Quantitative evaluation based on Noise Equivalent Temperature Difference (NETD) measurements showed that the proposed 16-frame processing achieves an NETD of 0.0241 °C, comparable to a cooled infrared camera under the tested conditions. The proposed system provides a physically interpretable and engineering-oriented solution for robust daytime IRT, improving reflection tolerance and practical applicability in infrastructure inspection.