Research on Bridge Crack Detection with Neural Network Based Image Processing Methods

In bridge health monitoring, the detection and localization of surface defects are highly important for health condition evaluation. Due to the limitation of manual detection, it is easier to measure those defects in a more automatic way. Machine learning is a hot topic in the recent decade, and the contribution of Artificial Neural Network (ANN) is especially remarkable, which is the most widely used models of machine learning in the image-processing field. In this paper, we will discuss two ANN-based algorithms (Back propagation (BP) and Self-Organizing Maps (SOM)) and their applications for the recognition of surface defect on images taken from bridges. Moreover, a combined network algorithm with BP and SOM is designed in order to improve the performance in crack image segmentation, and analysis over this network is carried out specifically.