Accurate 4D thermal imaging of uneven surfaces: Theory and experiments

Abstract

Conventional IR thermography is basically a 3D imaging model: pixel position (2D) and temperature (1D). However, the surface roughness, which cannot be depicted by the 2D geometry imaging, may introduce significant error into the temperature measurement, especially in close range. Although there are some principle studies on this theme, quantitative tools for arbitrary-shaped surfaces is still lacking. In this work, we have developed an imaging system composed of a binocular camera using structured illumination and an IR camera to reconstruct 4D uneven surface thermal images. This 4D information includes the 3D geometry of the surface and the temperature mapping over it. Based on the highly accurate surface roughness image obtained, a meshed area calculation tool is used to establish a temperature self-correction module for rough/uneven surfaces, which include convex surface, concave surface or their combinations. The effective emissivity of convex as well as flat surface is easy to calculate, because is this case, there is no inter-radiation or reflection. On contrast, the determination of concave surface emissivity is still posed as a challenge. In this work, some theoretic gap was first filled and then the experimental verification was provided. The temperature-correction takes angle of view and emissivity calibration into account. The angle of view is relative to the position of the object point and the emissivity calibration is based on the 3D geometry of the uneven surfaces. An imaginary flat surface is used to provide an effective emissivity for a local roughness. Experimental results shows that without this temperatue-correction module, the deviations between IR thermography and thermocouple are from ∼2.35 °C (at 50 °C) to ∼4.96 °C (at 300 °C) and from ∼2.43 °C (at 50 °C) to ∼4.78 °C (at 300 °C), while after the correction, the deviations are from ∼0.12 °C (at 50 °C) to ∼0.20 °C (at 300 °C) and from ∼0.16 °C (at 50 °C) to ∼0.20 °C (at 300 °C). Utilizing the thermal imaging model for recorrection, the deviations are from ∼0.13 °C (at 50 °C) to ∼0.19 °C (at 300 °C) and from ∼0.16 °C (at 50 °C) to ∼0.19 °C (at 300 °C). This generalized solution paves the way for high-accuracy temperature measurement on complexly rough surface, especially for high temperature, at which the measuring deviation can be very large.