Pavement distress classification using neural networks

A novel approach of applying moment invariants and neural networks to analyze pavement images is presented in this paper. By calculating moment invariants from different types of distress, features are obtained. Then a backpropagation neural network is used to classify these features. This approach is illustrated using randomly selected sample of video images of real cracks. Based on these samples, the feasibility of using moment invariants and neural networks to classify different types of crack is proven.<<ETX>>