Computer vision-based classification of cracks on concrete bridges using machine learning techniques

Concrete crack is a significant indicator related to the durability and serviceability of concrete civil infrastructure such as dams, bridges and tunnels. Current inspection of concrete structures is based on manual visual operation, which is not effective in safety, cost and reliability. This research aims to address the problems in traditional inspection of concrete structures by proposing a novel automatic crack identification approach, which intelligently integrates both image processing and machine learning techniques. Through the crack-sensitive feature extraction and model self-learning, the proposed method has higher identification accuracy than conventional inspection method, which has been proved by the experimental verification.