Evaluation of the use of Cascade Detection algorithms based on Machine Learning for Crack Detection in asphalt Pavements
Pavement maintenance aims to provide optimal service conditions. To achieve this objective, it is necessary to efficiently carry out pavement monitoring and inspection activities. In this field, damage analysis of road infrastructure by means of image processing has emerged as a low-cost alternative to other, more sophisticated methods. In this research, a method of damage detection in asphalt pavements is developed by analyzing images obtained from an unmanned aerial vehicle and a smartphone using a cascade detection algorithm based on machine learning. The contributions of this research are its demonstration of the use of an algorithm developed mainly for face detection in a pavement damage detection context and its proposal of recommendations and a workflow for image acquisition and processing. To validate the detector and measure its accuracy, a data set was created consisting of images of real pavement damage cases. The results show that the cascade detection algorithm is capable of identifying cracks and potholes, achieving optimal accuracy in the analysis of real cases.
Keywords: pavement inspection; pavement cracks; cascade detector; machine learning (ML); unmanned aerial vehicles (UAVs)