Modelagem termográfica: uso de fotogrametria em imagens termográficas de edificações para visualização tridimensional de manifestações patológicas

Abstract

It has long been known about the existence of buildings defects that cannot be visibly identified early, such as detachment of ceramic cladding on facades, when on the outside of buildings, and infiltrations, when on the inside of them. These manifestations usually interfere with the thermal coefficient of the elements that make up the building, causing a discrepancy in the emission of infrared radiation, which can be measured in a thermogram. In order to better visualize and understand the set of pathological manifestations, with regard to temperature, an orthomosaic composed of several superimposed thermographic images results in a three-dimensional image on a thermal scale, which facilitates the identification and measurement of the analyzed defects. Through it, it is possible to navigate throughout the modeled surface, perceiving the colorimetric variations in an isolated or systemic way. For this purpose, in this study, images were captured through cameras with an infrared sensor, orthogonally to the recorded surface, so that they could be processed in Agisoft Photoscan, a photogrammetry software, allowing the comparison between the data obtained in the field with the model full-scale digital surface of the building to be studied. The capture took place with two thermal imaging cameras, FLIR C5 and FLIR A325 and with two UAVs, DJI Phantom 4 Advanced and DJI Mavic 2 Enterprise Advanced. After processing the images, a three-dimensional model of the building was obtained, keeping the dimensional properties in temperature scale, thus allowing an easy reading and understanding of the data to carry out the search for buildings defects. The results obtained in the comparison between the measurements of the facades of the building in this study, acquired on the project and on the IR and RGB orthomosaics, indicate that photogrammetric processing on thermographic images was possible, resulting in a model with sufficient scale accuracy to perform measurements of distances and areas of defects, with an average accuracy of 97.37% in the center of the orthomosaic IR and 98.46% in the center of the orthomosaic RGB.