Visual defect recognition and location for pulsed thermography images based on defect-background contrast analysis

The basic principle of infrared thermography is that the temperature of intact regions is different from that of the defect regions because of thermal conductivity difference caused by discontinuity in the material inner structures. However, the defect information is often masked by measurement noise and non-uniform backgrounds. The thermal image sequence of testing material often reaches hundreds of frames, which also increases the difficulty of detection. Previous work mostly focuses on defect signal enhancement on specified image frames and it requires manual analysis for preliminary selection and further defect recognition. Besides, defect information is analysed at pixel-wise level, which ignores spatial relationships between pixels. To solve the above-mentioned problems, this paper proposes a defect-background contrast analysis for automatic defect recognition and location using region-based defect features. The efficacy of the proposed method is verified through an experiment based on carbon fiber reinforced polymer composites.