Development of a Method for Evaluating Pavement Texture Using Texture Features and Ensemble Learning

Key Points

• The main aim is to develop an accurate method for measuring pavement texture using advanced machine learning techniques.
• Utilized vehicle-mounted cameras for data collection.
• Compared various deep learning and machine learning models.
• Developed an ensemble learning model combining Random Forest and LightGBM.
• Applied the model to estimate average pavement texture depth (MPD) in a large parking lot.
• Achieved an R2 value of 0.85 indicating strong correlation with actual measurements.
• Obtained a root mean square error (RMSE) of 0.14mm, matching traditional methods' precision.
• Demonstrated improved accuracy in pavement texture measurement using the proposed method.