基于道路材料约束对抗神经网络的探地雷达数据生成方法
By standardizing and fusing ground-penetrating radar B-SCAN data and road material parameters, and combining adversarial training with a three-level hierarchical generator and a two-dimensional discriminator, the problems of hierarchical features and material constraints in ground-penetrating radar data generation in traditional models are solved. This achieves higher accuracy and applicability in data generation and improves the intelligence level of road damage detection.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CHANGSHA UNIVERSITY OF SCIENCE AND TECHNOLOGY
- Filing Date
- 2026-04-08
- Publication Date
- 2026-07-17
AI Technical Summary
Traditional generative adversarial neural network models do not consider hierarchical features in the generation of ground-penetrating radar B-SCAN data, and do not construct effective constraint mechanisms for road material parameters. This results in poor adaptability of the generated data to the actual road geological background, difficulty in reproducing the differences in radar echo signal response, and low model training stability and generated data reliability.
By acquiring raw ground-penetrating radar B-SCAN data and road material parameters from the same road detection area, standardized preprocessing is performed, and then channel splicing is performed to construct multimodal features. Random noise and geological constraint vectors are fused and input into a three-level hierarchical generator. Combined with a two-dimensional discriminator, adversarial training is carried out to generate ground-penetrating radar B-SCAN single-channel data step by step.
It has achieved hierarchical and precise generation of ground-penetrating radar data, improved the accuracy and applicability of road detection, generated more realistic and reasonable data, enriched the ground-penetrating radar data of road damage areas, and improved the level of intelligent detection.
Smart Images
Figure CN121997278B_ABST