Concrete Structure Defect Detection Using You Only Look Once Version 5 with AlexNet

Concrete structures are made from cement, water, and admixtures, which are used in buildings, water management systems, and foundations. There are many face pressing issues such as corrosion, cracking, and chemical attacks, requiring innovative solutions for sustainable infrastructure. Various methods have been proposed previously to recognize the defects of concrete structures. To overcome these problems, a Deep Learning (DL) model of You Only Look Once version 5 with Alex-Krizhevsky Network (YOLOv5- AlexNet) is proposed for detecting the defects of concrete structures. Primarily, a Surface Crack Detection (SCD) dataset is considered as an input for recognizing the defects of concrete structures. For preprocessing, image resizing and normalization, data augmentation, contrast enhancement, image filtering, and image binarization are considered for efficient processing, enhancing robustness, improving crack visibility, edge detection, and to simplify crack detection. Then for feature extraction, U-shaped Network (U-Net) is employed to reduce noise sensitivity, and to fine-tune the model. A YOLOv5-AlexNet is considered for classification to improve performance and generalizability, and random search for hyperparameter tuning. The proposed YOLOv5-AlexNet model attained higher performance metrics such as accuracy, precision, recall, and f1 score of 99.57%, 98.23%, 98.65%, and 99.86%, when compared to other existing methods like Convolutional Neural Network (CNN).