Color recognition with a camera: a supervised algorithm for classification

This paper presents a color recognition algorithm using neuronal techniques, applied to the characterization of thermal paints. The color of such a coating changes with temperature, allowing temperature maps to be established on gas turbine engine components for instance. After a formulation of the classification algorithm, performances are evaluated on a thermal paint samples set. The influence of lighting is studied. The results are compared with those obtained by more classical segmentation algorithms. A temperature profile on a test piece is finally achieved.