IMPROVED DETECTION OF THERMAL ANOMALIES IN BUILDING FACADES

Abstract

The aging of building facades has led to various defects on facades. The detection of latent defects is an essential part of facade inspection. Infrared thermography is adopted to identify areas with elevated temperatures during daytime, which are indicative of thermal anomalies that may be caused by subsurface defects. This study proposes a framework of thermal anomaly detection that combines plane segmentation and a revised region growing algorithm. Plane segmentation is implemented to determine the region of interest. The revised region growing algorithm selects seed points using 0.9 quantile and 32×32 window, and anomalies are detected through a thresholding method inside automatically adjusted local regions. A method of automatically excluding the influences of the edge effects is developed. The range of the edge effects is calculated by the moving average method and the linearity of the temperature contour is measured by the least absolute residuals.