Abstract: Subgrade loosening is a common hidden disease in asphalt road surfaces. Failure to promptly address it can lead to severe structural damage to the road surface, affecting road safety and increasing maintenance costs. Accurately identifying subgrade loosening's location, size, and water accumulation level through 3D-radar imaging is crucial for early detection and treatment. This study conducted quantitative investigations using forward modeling with 3D-radar to analyze radar responses to subgrade loosening with different sizes and moisture content. Validation involved simulating loose embeddings, extracting features from radar images via binarization, and taking core samples from existing road sections within the project area to verify typical image accuracy. Results show that after applying the feature extraction method to 3D-radar images, subgrade loosening's features become clear. Crescent-shaped visible regions appear in loosened subgrade areas, and dot-block patterns emerge beneath water-rich loosened subgrade areas. Subgrade loosening's radar characteristics correlate with size and moisture content, with larger diseases displaying smaller parabolic curvature features. Higher moisture content results in wider, bright strip patterns in loosened subgrade areas.