@article{article_1628600, title={DETECTION AND PREDICTION OF CONCRETE CRACKS USING DEEP LEARNING-BASED IMAGE PROCESSING METHODS FOR QUALITY CONTROL}, journal={Konya Journal of Engineering Sciences}, volume={13}, pages={1158–1174}, year={2025}, DOI={10.36306/konjes.1628600}, author={Karataş, İbrahim}, keywords={Concrete Cracks, Quality Control, Vision Technics, Pre-train Models, Voting Ensemble Model}, abstract={One of the most critical defects in the quality control process of concrete elements is the detection of cracks. Furthermore, cracks are among the most significant indicators affecting concrete strength. Manual crack detection presents numerous disadvantages in terms of time, labor, cost, high error probability, and practical implementation challenges. Therefore, this study aims to detect cracks on concrete surfaces using vision techniques and automatically predict them using deep learning methods. Images classified as crack and non-crack, selected from a dataset obtained from the literature, were initially analyzed using Canny and Threshold methods. Subsequently, analyses were conducted using a novel voting ensemble model that combines deep learning models such as VGG16, ResNet50, Xception, and MobileNet. According to the results, cracks were successfully detected using vision techniques, and the proposed voting ensemble model achieved an accuracy value of 99.75% with a loss value of 0.00618. The findings demonstrate that automated quality control of concrete surfaces specifically for cracks can be performed with high accuracy.}, number={4}, publisher={Konya Technical University}