Emissivity correction using spectrum correlation of infrared and visible images

Thermography is a typical sensing approach used for non-destructive testing and evaluation (NDT&E). However, the varying surface emissivity of an object leads to illusory temperature inhomogeneity which results in influences on defect detection. This paper proposes a new technique to correct the influence of the surface's varying emissivity of an object in active thermography. Two cameras operating at different spectra are used to capture infrared and visible images simultaneously. Although the physics behind infrared and optical imaging are very different, a close spectrum correlation of two images is identified. An invariant coefficient feature has been estimated for an emissivity correction of infrared images with suggested algorithm. The basic hypothesis is that the reflectance correlation is proposed to predict surface emissivity of an object with respect to wavelength. Experimental validation results show that after correction infrared images are looking like more homogeneous and independent of emissivity. It has been tested for a partially painted steel and rail samples with different known emissivity. Comparative analysis demonstrates its promising capability for accurate mapping of thermal patterns and defect evaluation in thermography NDT&E.