Open Access

ARTICLE

UAV-Deep Learning-Based Approach in Civil Structural Diagnosis

Wael A. Altabey^*
Department of Mechanical Engineering, Faculty of Engineering, Alexandria University, Alexandria, Egypt
* Corresponding Author: Wael A. Altabey. Email:
(This article belongs to the Special Issue: Modern Inverse Analysis Approaches for Structural Diagnosis and Parameter Identifications)

Structural Durability & Health Monitoring https://doi.org/10.32604/sdhm.2026.079495

Received 22 January 2026; Accepted 06 May 2026; Published online 22 July 2026

Abstract

The goal of this paper is to improve the monitoring of civil structures when we pair unmanned aerial vehicles (UAVs) technology with the current proposed algorithm, particularly to identify cracks in concrete structures. Typically, the current UAV methods are more about creating state maps of these structures, but they struggle with the impact of the drone’s movement on crack detection accuracy. This presents challenges for using intelligent systems for concrete crack detection. The current approach combines advanced technologies with a network of high-definition cameras mounted on inspection UAV systems and distributed in different parts of the structure’s surface, and smart processing methods to keep tabs on structure health accurately and in real-time. By merging a novel Convolutional Neural Network (CNN) with a specialized technique called the Crack Contour Network (CCN) to extract features, we boost the accuracy and reliability of our assessments for target identification, the type, location, and dimensions of cracks, including their borders. After collecting those images using a high-definition camera-based UAV system, they preprocessed the images by cropping and segmenting the crack images, standardizing their size, adjusting the contrast, enhancing the image data to boost the sample count, and creating a database with these labeled samples. Once that’s done, use this database for training the CNN-CCN, which was built and tuned to the parameters of the network. Finally, the network tested that’s been trained, and the algorithm will output the detection results for the cracks. An algorithm CNN-CCN uses metrics like Intersection over Union (IoU), accuracy rates, regression rates, and F1-score to evaluate its performance in crack identification and determining the crack’s contour area. The results indicate that the proposed CNN-CCN network performs well, achieving high IoU for crack detection, accuracy rates of 96.43%, regression rates of 93.77%, and an F1-score of 91.65%, alongside a training time of 75 s and a model size of 199 KB. A deep analysis of the confusion matrix and a side-by-side look at predicted and actual conditions highlight how well the model can tell apart different traditional algorithms in the literature. This technique allows for automatic high-dimensional extraction and complex features from infrastructures, leading to reduced interference and errors in detection, which enhances the clarity and overall effectiveness of civil structure assessments.

---

Keywords

Deep learning; civil structural diagnosis; UAVs; convolutional neural network (CNN); crack identification; crack contour network (CCN)

---