An Automated Assessment Technique for Pavement Defects Using a Laser Scanner and Deep Machine Learning

Roads are vital arteries and main links between and within cities. They are considered the main auxiliary factor in shortening travel time and achieving users' comfort and safety. Governments strive to provide ideal conditions on the roads to achieve the highest levels of satisfaction, which are reflected in the quality of rides provided. Despite the variety of monitoring and evaluation methods, achieving the best and most accurate diagnosis of the condition of the roads and determining the severity of defects and appropriate and rapid maintenance methods are still lacking. This study aims to monitor and evaluate the state of some roads in Aswan City, Egypt, to identify defects and address them promptly. To achieve this goal, a laser scanner was used to evaluate pavement conditions by measuring the coordinates of the road surface and determining the differences in the measured values on the three axes. A built-in camera was also used in the laser device to monitor the type and severity of defects and match them with the measurements of the laser scanner device. Finally, a deep machine learning system, including LSTM, GRU, RF, SVM, and DT, was used to identify and classify the type and severity of defects. The prediction models showed significant accuracy with about 93%, 91%, 85%, 84%, and 82%, respectively.

Doi: 10.28991/CEJ-2025-011-03-015

Full Text: PDF