Automated framework for evaluating asphalt pavement construction using UAV imagery

Abstract

Thermal uniformity and in-place density are key quality assurance factors affecting freshly placed asphalt pavement (referred to as mat) performance. Existing techniques like Paver Mounted Thermal Profilers and Intelligent Compaction Technologies quantify thermal non-uniformities and density differentials. In this paper, an alternative protocol was developed using unmanned aerial vehicle (UAV) assisted aerial infrared (IR) image data. A deep-learning object detection model was developed to identify the location of the paved mat and rollers in each thermal image using the YOLOv8 model. The developed framework was used for various sites visited in 2023 and 2024. Three quantification metrics for thermal segregation—Differential Range Statistic (DRS), Thermal Segregation Index (TSI), and Mat Temperature Differential Matrix (MDM)—are compared across all sites’ thermal images. A case study based on data from an experimental site is presented, and two lanes were monitored for roller movements. Compaction metrics such as the overall roller pass counts, roller speed, and insufficient roller passes based on the bottom quantile data were obtained from the developed protocol. The presented framework successfully identifies and maps the roller movements while calculating the thermal segregation and compaction metrics. When processed in the field during construction, the metrics can give near-real-time insights into the mat’s non-uniformities during paving. This developed protocol can provide actionable feedback to the contractors/agencies on the job site and help improve overall consistency in the quality of construction.