Automated crack detection in flexible pavements on urban roads to ensure service life

Abstract

The objective of this study was to use artificial intelligence-based image analysis to detect cracks in flexible pavements within the urban area of the La Banda de Shilcayo district (Tarapoto). This is a non-experimental study with a descriptive-comparative design. The method involved collecting photographic images, which were then processed using the Python programming language and the YOLO algorithm to a model was developed that allows for the identification and quantification in millimeters of cracks in the pavement through image visualization. These cracks were categorized into analysis units according to the PCI method, thereby not only providing reliable and rapid information but also reducing the costs and time required for technical staff sent to the field to gather this same information. The results obtained have been compared with the information collected in the field, yielding an average error margin of 1 cm to 2 cm difference from the information provided by the developed model, specifically regarding the length of the cracks. Using the PCI method, it has been possible to determine the condition of the pavement and identify which areas require routine maintenance, recurring maintenance, or periodic maintenance. Furthermore, the method has demonstrated a reduction in manual labor and time spent obtaining real-time information on road conditions, which has resulted in economic benefits for road administrators. However, beyond economic benefits, the importance of keeping this information up-to-date lies in enabling timely interventions that facilitate the services provided to citizens.