A Method of Concrete Surface Crack Detection Using an Improved Convolutional Neural Network (CNN) Model

This essay spotlights concrete crack detection in infrastructure maintenance, highlighting its importance for structural integrity, cost-effectiveness, and eco-consciousness. It delves into various detection methods and introduces an improved VGG-16-based deep learning model with batch normalization, P-ReLU activation, and Adam optimization for better training outcomes. Through experiments on the MendeleyData-CrackDetection dataset, the enhanced model outperforms the original. This study underscores the significance of hyperparameter optimization and algorithm choice in deep learning.