Deep learning and TOPSIS-based multi-criteria decision-making framework for urban road defect detection and sustainable maintenance planning.

In most cities, especially large ones, one of the most significant factors influencing the Quality of Life (QoL) is the urban and public transportation networks. In order to guarantee traffic safety, lower maintenance costs, and promote the sustainability of contemporary cities, urban road infrastructure must be maintained. Conventional techniques for identifying surface flaws in roads are frequently time-consuming, labor-intensive, and unable to satisfy the real-time requirements of intelligent urban management. However, most of the currently available models focus only on the detection of surface flaws without estimating the severity or linking the outcomes to actionable maintenance priorities. Such a solution makes their usage limited for real-time urban management and sustainability-based decision-making. To overcome these limitations, the present study has proposed an improved YOLOv8 model with an additional secondary head for severity classification and incorporated Multi-Criteria Decision-Making through TOPSIS. The new dual-framework model enables the simultaneous performance of detection, estimation of severity, and data-driven prioritization for maintenance planning, becoming a major leap beyond currently used one-dimensional methods of detection. The enhanced model presented in this research performs better than the original YOLOv8 with MCDM, according to results from experimental on the road defect detection dataset. Results show that the suggested framework can achieve a viable short-term strategy without sacrificing long-term effectiveness. It promotes proactive infrastructure upkeep, strengthens the resilience of urban mobility, and aids in the creation of more secure and sustainable transportation systems.