A Deep-Convolutional-Neural-Network-Based Semi-Supervised Learning Method for Anomaly Crack Detection

Crack detection plays a pivotal role in structural health monitoring. Deep convolutional neural networks (DCNN) provide way to achieve image classification efficiently and accurately due their powerful processing ability. In this paper, we propose semi-supervised learning method based on DCNN anomaly crack detection. the proposed method, training set for network only requires small number of normal (non-crack) images but can high accuracy. Moreover, trained model has strong robustness condition uneven illumination evident difference. The is applied walls, bridges pavements, results show that accuracy comes up 99.48%, 92.31% 97.57%, respectively. addition, features be visualized describe its working principle. This great potential practical engineering applications.