A UAV Photography–Based Detection Method for Defective Road Marking

AbstractMarkings are essential elements of roads and are important for driving safety enforcement. Traditional detection of marking defects relies on manual inspection, which is less effective. However, few studies could be found focusing on automatic detection of defective road markings. In this study, a detection method for defective road marking is proposed based on unmanned aerial vehicle (UAV) photography. Image acquisition based on UAV photography was discussed. The detailed processes of image preprocessing, road segmentation, marking extraction, and defective marking detection are presented. The proposed method was tested with images collected from highways, normal urban roads, and poorly maintained urban roads in Nanjing, China. Overall, the proposed method shows good robustness in defective marking detection. The precision and recall of marking extraction on highways and normal urban roads are over 93%, and the results on poorly maintained urban roads are about 76%, due to the poor road conditions. The precision of damage identification on lane markings and arrow markings were all greater than 90%. The proposed automatic process is capable of accurately and quantitatively detecting the sidelines, lane markings, and indicator markings.