An automatic pothole detection algorithm using pavement 3D data

Road pavements are subject to various forms of degradation compromising their functionality with negative effects on safety. For assuring the highest quality, all the distresses have to be properly identified and quantified by road administrators. For increasing efficiency and reducing costs and times of surveys, several innovative methods to detect, classify and measure surface distresses were proposed, with variable results. In this context, the authors propose an algorithm for automated pothole detection through the processing of 3D data of pavement surfaces, acquired using an innovative high-performance equipment. The algorithm, derived from computer vision, is able of identifying potholes in road sections, assuring a reliable estimation of shape and severity, in terms not only of area, perimeter, but also depth, with practical benefits. The numerical results show the remarkable performance of the proposed algorithm, even compared to alternative traditional methodologies. In terms of Precision, Recall and F-Score, it assures mean values equal respectively to 89.75%, 92.95% 91.28%. Validation was also performed in terms of area error rate, with an average value of 5.15%, significantly lower than other approaches. Then, the algorithm represents a reliable alternative to traditional approaches and allows road administrators to derive data to optimize maintenance and road functionality.