講演情報

[Data_Collection-34]From Empirical to Intelligent: Optimization of Grading Standards for Pavement Distress in Automated Inspection

Wenruo Fan1, *Hui Wang1, Shengchuan Jiang2, Zhoucong Xu3,4 (1. Chongqing University (China), 2. University of Shanghai for Science and Technology (China), 3. Tongji University (China), 4. China Merchants Chongqing Communications Technology Research&Design Institute Co.,Ltd (China))

キーワード:

Pavement surface distress、Intelligent grading、Crack detection、Machine vision、Evaluation standard

Currently, the grading and evaluation system for road surface distress detection is primarily based on traditional manual surveys and empirical statistics. Although this system has accumulated valuable experience over long-term maintenance practices, its issues—such as strong subjectivity, low efficiency, and lack of standardization—are becoming increasingly apparent. Especially against the backdrop of rapidly developing automated inspection technologies, existing empirical standards struggle to meet the requirements of objectivity, consistency, and refined interpretation demanded by machine vision systems. This often leads to an oversimplified distress identification process and a significant gap between evaluation results and the complex, variable conditions of actual road surfaces.To address this core challenge, this study proposes an intelligent grading standard optimization framework tailored for automated inspection. The framework aims to facilitate a fundamental shift in the road surface distress evaluation system from “experience-driven” to “data- and intelligence-driven.” The research deeply integrates advanced computer vision technologies, particularly semantic segmentation and fine-grained classification algorithms, to construct a multidimensional intelligent analysis model capable of comprehensively interpreting crack morphology, width, density, and associated damages. By systematically learning from extensive annotated data, the model achieves precise quantitative extraction and semantic understanding of distress features, thereby transforming traditional qualitative and coarse empirical descriptions into objective, detailed, and computable quantitative indicators.The core contribution of this study lies in developing a set of adaptive grading standards that align with automated inspection technologies. These standards not only significantly improve the accuracy and consistency of crack identification and classification but also strike a critical balance between “the complex demands of real-world conditions” and “efficient automated processing.” The research outcomes are expected to provide a scientific, precise, and scalable intelligent decision-support tool for road maintenance management, promoting the digital and intelligent transformation of infrastructure maintenance.