Automated pavement crack detection is essential to a cost-effective pavement asset management system. Many automated crack detection algorithms (CDAs) have been developed, but they lack a standardized performance evaluation system, which is urgently needed. This paper presents a comprehensive, quantitative CDA performance evaluation system (CDA-PES) that includes: (1) an enhanced Hausdorff distance–based quantitative performance evaluation method; (2) a consistent data set designed with diverse pavement types and crack conditions that affect CDA performance; and (3) a multilevel, categorized, quantitative performance scoring and reporting method to critically assess the CDA performance and provide feedback to researchers. Six CDAs are used to demonstrate the proposed system. The CDA-PES is available on the internet. Transportation agencies can use CDA-PES to objectively evaluate and compare different detection algorithms, and researchers can use it to improve their algorithms. The CDA-PES provides a consistent, comprehensive, and quantitative performance evaluation system for advancing automated CDAs.
No takes yet. Share an insight, caveat, or question.
Tsai et al. (2017) studied this question.