Enhanced crack detection using GOA-UNet segmentation and Connected Component Analysis based aspect ratio estimation

Crack detection is critical for maintaining the safety and longevity of infra-structure, yet traditional methods often rely on manual inspection or basic image pro-cessing techniques, which are labor-intensive and prone to errors. To overcome these challenges, this paper proposes a novel crack detection approach using a GOA-Unet segmentation model. The Grasshopper Optimization Algorithm (GOA) is employed to optimize the hyperparameters of the U-Net architecture, improving detection accuracy by addressing issues such as small crack detection and false positives. Furthermore, Power Law Transformation is applied to enhance image contrast, ensuring that subtle cracks are more visible. The Generative Adversarial Network (GAN) is used for data augmentation, generating realistic crack patterns to increase the diversity of the training dataset. This augmentation helps the model generalize better across various crack types and imaging conditions. Additionally, Connected Component Analysis (CCA) is utilized for aspect ratio calculation, providing detailed geometric information about the detected cracks, which is critical for assessing their structural significance. The pro-posed GOA-Unet model achieves a crack detection accuracy of 98.5% across multiple datasets, outperforming conventional models in terms of precision and generalization. These findings demonstrate that the integration of GOA, Power Law Transformation, GAN-based augmentation, and CCA offers a robust solution for accurate and efficient crack detection in real-world application. • GOA automates U-Net hyperparameter tuning for optimized crack segmentation. • GAN-based augmentation addresses data scarcity, boosting model generalization. • Power Law Transformation enhances contrast, enabling fine crack detection. • Integrated CCA extracts crack geometry for quantitative severity assessment. • Attained 98.5% accuracy rate on challenging crack identification tasks.