Data quality assessment of automated pavement cracking measurements in Mississippi

Abstract

The reliability and cost-effectiveness of pavement maintenance and rehabilitation (M&R) decisions are highly dependent on the quality of pavement performance data which are regularly collected and stored in the Pavement Management System (PMS) of a state highway agency. Frequently, pavement performance evaluation relies on the precise and accurate measurements of the surface conditions of pavements. This study aims to statistically check the accuracy and precision of using a state of practice automated technology in place of the manual or semi-automated (semi-manual) rating method to collect pavement cracking data. In the study, the classification evaluation indicators were first employed to quantitively evaluate the two kinds of errors which may occur during the automated crack detection process. Second, the systemic errors of the automated method were quantitively evaluated by a t-test to check the accuracy of the automated method, and the variation pattern of the systemic errors was verified with the Pearson correlation test. Finally, the random errors of the automated method were quantitatively evaluated with the Levene’s test to check the precision of the automated method. The study results indicated that the state of practice automated survey method is still not yet ready to provide reliable pavement cracking measurement data. It seemed the automated method in the study could not accurately detect pavement cracking features from the background and an under-estimation tendency of cracking measurement was observed. The automated data also showed varying measurement performance when surveying different types of pavement cracking. While the collected transverse cracking data were relatively precise, obvious measurement errors were present in longitudinal cracking data. In addition, the systemic errors in the automated data were highly correlated with the surveyed pavement distress condition. A greater systemic error was frequently associated with a more deteriorated pavement condition.