Browsing by Author "Serafim, Barbara Gomes"
Now showing 1 - 1 of 1
Results Per Page
Sort Options
- Automatic detection of pathologies in buildings from thermal imagesPublication . Serafim, Barbara Gomes; Almeida, Ricardo Manuel dos Santos Ferreira de; Mazer, Wellington; Barreira, EvaThe application of thermography, as a non-destructive technique to evaluate pathologies in buildings where destructive tests are unfeasible or difficult to perform, stands out as a valuable tool with high potential. InfraRed Thermography (IRT) offers the ability to identify structural pathologies and abnormalities by inspecting thermal radiation from building surfaces. While widespread, the applicability of IRT is generally constrained by the quality of the resultant thermal images as well as their interpretative subjectivity. The aim of this research is to circumvent these shortcomings by developing an automatic system for the identification of construction pathologies using thermal images, which would make building inspections more accurate and efficient. The approach employed in this work involved the collection and categorization of a database of thermal images, which served as the foundation for developing the automatic detection tool. The research process was structured into several primary phases, including problem definition, data collection, image processing, implementation of software, and assessment. The results of this study confirmed the instrument to be useful in detecting and mapping construction irregularities. The presence of noise in the thermal images, nonetheless, posed a significant threat to the accuracy of the detection process. Comparative studies indicated that color detection methods outperformed percentile-based methods in terms of accuracy and reliability. The image processing functions were highlighted, although drawbacks like fixed resolution and potential memory utilization issues were noted. In conclusion, this research contributes to automated pathology detection in building diagnostics using infrared thermography. The developed tool can potentially assist technicians with a more efficient and objective means of interpreting thermal images. The tool's performance should be further enhanced, with the inclusion of machine learning algorithms for improved accuracy, and the database populated to enable stronger neural network training. Additionally, the incorporation of a report features to categorize discovered pathologies by extent of damage would also make the tool more realistically functional.