The state of the road network contains both the infrastructure performance and the traffic operation state of the road network. There is a strong coupling between the decay of the infrastructure performance and the redistribution of the traffic flow on the road network. In this paper, a dynamic road network description model is proposed to apply to the couple simulation of highway network infrastructure performance and traffic state. First, a road network description model is constructed by associating the highway network topology with state attributes. The topology contains traffic information and is dynamically editable. Then, a dynamic road network model is proposed that can dynamically represent the changes in local connectivity relationships caused by traffic control, such as lane/ramp closures and turning restrictions in actual roads due to construction operations and access to the state of multi-scale spatio-temporal road networks. It overcomes the defects of the existing road network model, which is difficult to apply to the analysis of service performance and traffic state of the road network in different periods. Finally, the application of the dynamic road network model in the highway network coupled simulation system (HNCS) is completed, which provides a method for improving the efficiency and accuracy of large-scale highway network traffic simulation and highway infrastructure performance prediction.

To reduce the reliance of existing mainstream vision-based crack detection algorithms on annotated datasets, this paper presents an unsupervised semantic segmentation-based approach. The proposed method first employs the Felzenszwalb-Huttenlocher algorithm for pre-segmenting the image, generating superpixels. Subsequently, a with an autoencoder model is designed to progressively approximate the superpixel segmentation results, and the optimal model is obtained by Bayesian optimization. Through comparative experiments with existing algorithms, it has been demonstrated that the proposed method performance comparable to supervised algorithms, even without the need for labeled data. As a result, the deployment complexity of the algorithm is significantly reduced, while expanding its applicability.