Automatic Pothole Detection Using ISO Cluster Unsupervised Classification

Potholes are structural failures on the road surface in the form of irregular circles on the pavement or asphalt surface of roads as a result of distressed roads due to traffic flow. The aim of this study is to identify potholes’ geometry using UAV quad copter for the purpose of road maintenance. This research consisted of four main phases: site reconnaissance and planning, data acquisition, data processing, and result and analysis. This research demonstrated the capability of using UAV images for road pavement data collection. Data were determined using two measurement types: the area and perimeter of potholes. This study showed that data on potholes taken from 8 m altitude gave accuracy about 0.022 m2 with a standard deviation of +0.026 m2. The accuracy of data extracted on potholes from 10 m altitude is 0.034 m2 with a standard deviation of +0.056 m2. The extracted pothole’s perimeter accuracy at 8 m altitude 0.021 m with a standard deviation of +0.054 m. The extracted pothole perimeter for 10 m altitude is 0.067 m with a standard deviation of +0.160 m. The data proved that extracted data on potholes from a UAV platform of 8 m altitude was most accurate compared to data extracted at 10 m altitude. In conclusion, the method used in this study can improve the road maintenance system implemented in the country through more accurate measurements of road surface distress.