An integrated vision material intelligence framework for infrastructure condition assessment and asset valuation

Abstract

The deterioration of civil infrastructure has increased the demand for data-driven approaches that support objective condition assessment and maintenance planning. Existing methods typically address visual defect detection, material strength prediction, and infrastructure assessment independently, limiting their ability to provide integrated engineering insights. This study proposes a unified infrastructure condition-assessment framework that combines computer vision, material strength prediction, and analytical condition assessment within a single workflow. The visual inspection module, Hierarchical Edge-Aware Residual Network (HEAR-Net), incorporates an Edge-Gated Activation mechanism to enhance boundary-sensitive crack detection under heterogeneous surface conditions. On the independent test set, HEAR-Net achieved an accuracy of 99.90%, precision of 99.90%, recall of 99.90%, specificity of 99.90%, F1-score of 99.90%, and an ROC-AUC of 0.999993. Material property estimation was incorporated as a complementary component of the proposed framework, where a regression-based model estimated concrete compressive strength from physicochemical composition variables, achieving an R 2 of 0.84 on the evaluated dataset. The outputs of the visual inspection and material property estimation components were subsequently integrated within an analytical workflow to demonstrate infrastructure condition assessment. Overall, the proposed framework demonstrates the feasibility of integrating automated visual inspection with complementary material property estimation in a unified infrastructure evaluation workflow, while the analytical condition assessment is presented as an illustrative demonstration requiring further validation on real-world infrastructure data.