Deep Learning model for crack detection in thermal images

Abstract

Automated crack detection in structures is crucial to ensure the durability and safety of buildings, especially in regions subject to thermal variations. This study proposes a Deep Learning model based on Convolutional Neural Networks (CNNs) for detecting and classifying cracks in thermal images of construction elements. A balanced dataset with 128 thermal images was used, expanded to 2000 samples via data augmentation techniques, such as adding noise and rotation. The CNN architecture included convolutional layers, pooling, propuot and the Sigmoid activation function for binary classification.