AMFFF: Road Damage Detection Method With Adaptive Multi-Frequency Feature Fusion

Road damage detection is a critical component of road maintenance and safety management. However, traditional methods often suffer from limitations in detection accuracy and robustness, especially in complex environments. To address this challenge, this paper proposes a novel road damage detection method based on wavelet transform and adaptive multi-frequency feature fusion, namely adaptive multi-frequency feature fusion (AMFFF). By leveraging wavelet transform, the AMFFF method performs multi-scale decomposition of road image signals to extract target features across different frequency components, thereby capturing fine details of road defects at various scales. Furthermore, in the AMFFF, an adaptive feature fusion strategy is designed to dynamically integrate features from different frequency components with the original features, significantly enhancing the discriminative feature representation of defect targets. Experimental results demonstrate that the proposed method effectively improves detection accuracy and robustness, particularly under challenging conditions such as complex lighting, shadows, and texture interference. Compared with existing approaches, the proposed AMFFF road damage detection method achieves significant improvements in detection accuracy and recall rate, offering a new perspective for the advancement of road damage detection technology.