A Noise-Cancellation Pipeline for GPR-Based Asphalt Pavement Compaction Evaluation

During asphalt pavement construction, compaction degree is a key indicator of quality control, directly affecting pavement service life and long-term performance. When mounted on a roller, air-coupled ground-penetrating radar (GPR) enables real-time pavement compaction evaluation due to its high efficiency, continuous measurement, and large spatial coverage. However, under practical construction conditions, multiple factors jointly affect the propagation and reflection of electromagnetic waves at the pavement surface. Surface moisture from water sprayed onto roller drums, as well as antenna height variations induced by roller vibration, can cause significant fluctuations in reflection amplitude, thereby reducing the accuracy and stability of GPR-based density predictions.

This study develops a time-domain signal correction framework to improve the accuracy of GPR-based density predictions during pavement compaction. The framework was designed to support automated processing by extracting the pavement surface reflection from full GPR signals and mitigating amplitude distortions induced by construction-related disturbances. Specifically, a semi-blind source separation method based on independent component analysis (ICA) was employed to remove surface moisture–related electromagnetic interference. At the same time, an electromagnetic-empirical model relating antenna height to reflection amplitude was introduced to compensate for vibration-induced variations in antenna height. By jointly accounting for these coupled effects within a unified correction strategy, the proposed framework recovered pavement surface reflections representative of dry conditions at a reference height, thereby enhancing the stability and reliability of GPR-based density estimation.

The proposed framework is validated through FDTD-based numerical simulations and field experiments. The results demonstrate that surface moisture effects and roller-induced antenna height variations can be effectively corrected, whether acting individually or in combination, allowing pavement surface reflection amplitudes to be recovered to a dry state at a standard antenna height. This work provides a practical basis for developing real-time GPR-based pavement compaction evaluation methods under complex construction conditions.