Machine Learning and Thermography Applied to the Detection and Classification of Cracks in Building (2212.14730v1)

Abstract: Due to the environmental impacts caused by the construction industry, repurposing existing buildings and making them more energy-efficient has become a high-priority issue. However, a legitimate concern of land developers is associated with the buildings' state of conservation. For that reason, infrared thermography has been used as a powerful tool to characterize these buildings' state of conservation by detecting pathologies, such as cracks and humidity. Thermal cameras detect the radiation emitted by any material and translate it into temperature-color-coded images. Abnormal temperature changes may indicate the presence of pathologies, however, reading thermal images might not be quite simple. This research project aims to combine infrared thermography and ML to help stakeholders determine the viability of reusing existing buildings by identifying their pathologies and defects more efficiently and accurately. In this particular phase of this research project, we've used an image classification machine learning model of Convolutional Neural Networks (DCNN) to differentiate three levels of cracks in one particular building. The model's accuracy was compared between the MSX and thermal images acquired from two distinct thermal cameras and fused images (formed through multisource information) to test the influence of the input data and network on the detection results.

• The paper presents a CNN-based framework that integrates thermal imaging and machine learning to detect and classify building cracks with over 96% accuracy.
• It employs dual thermal cameras and image fusion techniques to precisely measure crack severity based on defined temperature thresholds (ΔT values).
• The research paves the way for automated structural diagnostics, enhancing energy efficiency and adaptive reuse in the built environment.

Machine Learning and Thermography Applied to Crack Detection in Buildings

The paper "Machine Learning and Thermography Applied to the Detection and Classification of Cracks in Buildings" by Angela Busheska et al. presents a method integrating machine learning with infrared thermography to improve the detection and classification of structural pathologies in buildings. The overarching goal is to enhance the adaptive reuse of existing structures, thereby contributing to energy efficiency and sustainability in the built environment.

Methodology

The authors employ deep learning through a Convolutional Neural Network (CNN) to analyze thermal images and identify cracks at varying severity levels in buildings. The paper utilizes images captured with two different thermal cameras: HTI HT-18 and FLIR One Pro. These devices provide data that enable the classification of crack severity based on temperature differentials (ΔT) observed between the crack and its surrounding material. This approach is quantified as follows:

• Level 1: ΔT < 2°C
• Level 2: 2°C ≤ ΔT < 4°C
• Level 3: ΔT ≥ 4°C

The data preprocessing involves fusing thermal and conventional images to enhance feature recognition during training. The input images, resized for consistency, undergo noise reduction and sharpening for improved analysis. The dataset is split into training (60%), validation (20%), and test (20%) sets, ensuring robust model evaluation.

Results

The CNN model demonstrates high efficacy in crack detection across varying severity levels, achieving an accuracy surpassing 96%. Notably, the FLIR MSX images exhibit superior performance metrics, including precision, recall, and F1 score, compared to others. However, the detection of the lowest severity level cracks showed some limitations due to potential misidentification with the wall's structure. The research outcomes suggest significant promise in automating the crack detection process, offering reliable diagnostic support for building reuse projects.

Implications and Future Work

This research carries notable implications for both the practical implementation in building management and the theoretical advancement of image-based defect detection. By coupling thermography with machine learning, the paper establishes a framework that aids in the efficient diagnosis of building pathologies, potentially influencing policy and industry standards concerning structural maintenance and adaptive reuse.

Future work can explore the integration of thermal imaging drones to further expedite on-site data collection, enhancing accessibility and reducing human-induced biases in crack detection. This integration could also align with the vision of smart cities, where real-time data acquisition and analysis through AI-driven solutions facilitate better management of the urban built environment.

The paper highlights the increasing relevance of leveraging advanced technologies to address sustainability challenges within the construction sector. It sets a foundation for expanded research in applying AI to thermal imaging, focusing on automation and precision for structural assessment and conservation.

Conclusion

The integration of machine learning and thermographic techniques, as investigated in this paper, represents a significant step in advancing building pathology detection methods. This approach not only supports the adaptive reuse and conservation of architectural heritage but also offers a scalable solution for enhancing the energy efficiency of existing buildings. The research thus provides a vital contribution to the ongoing efforts of reducing the environmental impact of the construction industry.