The remote sensing method for large-scale asphalt pavement aging assessment with automated sample generation and deep...

Area of Science:

• Remote Sensing
• Deep Learning
• Pavement Engineering

Background:

• Asphalt pavement aging assessment is crucial for road maintenance and traffic safety.
• Traditional manual surveys are inefficient and time-consuming for large-scale monitoring.

Purpose of the Study:

• To develop an innovative framework for rapid and accurate assessment of asphalt pavement aging over large areas.
• To integrate multi-endmember mixed pixel unmixing, automated sample generation, and deep learning for pavement aging evaluation.

Main Methods:

• Utilized WorldView-3 remote sensing data for sample generation.
• Employed multi-endmember spectral unmixing and neighborhood filtering to create high-quality training and validation samples.
• Applied a one-dimensional convolutional neural network (1D-CNN) with an unsupervised zero-shot transfer approach for model training and inference.

Main Results:

• Achieved high classification accuracy (95.95%) and Kappa coefficients (0.9459) in study area-I.
• Demonstrated good performance in study area-II with 89.70% accuracy and 0.8628 Kappa coefficient.
• Validated the effectiveness of the proposed framework for large-area asphalt pavement aging assessment.

Conclusions:

• The integrated framework provides an effective solution for rapid asphalt pavement aging assessment.
• Results support improved pavement maintenance decision-making and traffic safety warning systems.