A pixel-by-pixel reliability analysis of infrared thermography (IRT) for the detection of subsurface delamination

This work presents an improved method for characterizing the reliability of Infrared Thermography (IRT) and determining the optimum thermal contrast for the detection of delamination in concrete. Receiver operating characteristics (ROC) analysis in conjunction with image processing methods were used to analyze data from a fabricated and in-service bridge decks. Results of the study demonstrated the excellent capability of IRT to discriminate delaminated from sound concrete. This conclusion was reached because the nonparametric estimate of the area under the ROC curve (AUC) was larger than 80% with its 95% confidence for each group of images. The optimum thermal contrast range was determined to be 0.6∽0.8 °C. Accurately defining the actual condition of the test object is critical to implement the proposed analysis method.