EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Real‐Time Concrete Crack Detection and Classification Using Multi‐Stage YOLOv8n Framework: A New Approach

Siraj Qays Mahdi, Sadik Kamel Gharghan, Ammar Hussein Mutlag, Hasan Kahtan

The Journal of Engineering · 2026-01 · 2026 (1)

DOI: 10.1049/tje2.70208 ↗

Kurz gefasst

Fachartikel in der Zeitschrift The Journal of Engineering, online erschienen am 15.07.2026. Verfasst von Siraj Qays Mahdi und 3 weiteren Autor:innen. Die Arbeit nennt 34 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Cracks in concrete surfaces are a major factor in the deterioration of concrete structures, such as bridges, roads and buildings. Therefore, early detection of cracks is crucial for assessing the integrity of concrete structures. To overcome the limitations of manual inspection, image processing techniques, particularly deep learning (DL) algorithms, have been explored over the past two years. In this paper, an accurate, real‐time image processing system was implemented to detect and classify cracks in concrete buildings into four categories: no cracks, minor cracks, moderate cracks and severe cracks. This was achieved using the YOLOv8n deep learning algorithm. The methodology for real‐time crack detection and classification in concrete buildings involves applying two models: the first model (model_1), based on YOLOv8n, is trained to detect the region of interest (ROI) of cracks and exclude potential defective areas in concrete images. The second model (model_2), based on YOLOv8n, is then trained to classify concrete images into four main categories. The dataset, containing both normal and abnormal images (with different cracks), was collected from Kaggle and from real images of our college buildings. The results of the image processing system based on the deep learning algorithm YOLOv8n showed that the first model (i.e., the crack detection model) was highly effective at accurately identifying crack regions, with a mean average precision (mAP50) of 0.50 and an intersection over union (IoU) exceeding 99%. In contrast, the second model (i.e., the crack classification model) showed promising classification results, with mAP50 of 95.6%, recall of 98% and F1 score of 91%. Both models are used in conjunction with a Web camera to detect cracks in real‐world environments.

Zitieren

APA

Mahdi, S. Q., Gharghan, S. K., Mutlag, A. H., & Kahtan, H. (2026). Real‐Time Concrete Crack Detection and Classification Using Multi‐Stage YOLOv8n Framework: A New Approach. The Journal of Engineering, 2026(1). https://doi.org/10.1049/tje2.70208

Vancouver

Mahdi SQ, Gharghan SK, Mutlag AH, Kahtan H. Real‐Time Concrete Crack Detection and Classification Using Multi‐Stage YOLOv8n Framework: A New Approach. The Journal of Engineering. 2026;2026(1). doi:10.1049/tje2.70208

BibTeX
@article{Mahdi2026,
  author = {Mahdi, Siraj Qays and Gharghan, Sadik Kamel and Mutlag, Ammar Hussein and Kahtan, Hasan},
  title = {Real‐Time Concrete Crack Detection and Classification Using Multi‐Stage YOLOv8n Framework: A New Approach},
  journal = {The Journal of Engineering},
  year = {2026},
  volume = {2026},
  number = {1},
  doi = {10.1049/tje2.70208}
}
Kostenlos starten →