BFMPF-Net: Bidirectional Frequency-Domain Modulation Progressive Fusion Network for Road Crack Segmentation
Wen Yang, Yingying Zheng, Hang Sun, Chao Liang, Lei Fang
AI summary
Problem
Existing frequency-domain methods neglect dynamic interactions between high- and low-frequency features, while standard encoder-decoder architectures suffer from progressive feature loss, limiting accurate crack detection in complex environments.
Approach
The proposed BFMPF-Net introduces a Bidirectional Frequency-domain Modulation Enhancement module to suppress noise and enhance crack edges, combined with a Progressive Guidance Fusion module that refines features stepwise using original low-resolution image priors.
Key results
- BFME module successfully attenuates background noise while preserving fine crack edge details.
- PGF module effectively mitigates feature dilution through coarse-to-fine progressive fusion.
- Outperforms nine state-of-the-art methods across Precision, Recall, F1 Score, and MIoU on three public datasets.
- Achieves optimal segmentation performance with the lowest computational cost and competitive parameter count.
Why it matters
Provides a highly accurate and computationally efficient solution for automated road infrastructure inspection and robotic maintenance systems.
Abstract
Recently, deep learning–based methods for road crack segmentation have achieved promising performance, particularly in robotic vision applications such as automated inspection and maintenance. However, most frequency-domain methods employ a decoupled processing strategy, overlooking the dynamic modulation mechanism between high- and low- frequency components, which constrains the model’s effec- tiveness in detecting cracks within complex environments. Moreover, existing methods suffer from low information fidelity during feature transmission, where critical encoder details are progressively lost in the decoder, making it difficult to reconstruct complete crack structures. To address these issues, we propose a Bidirectional Frequency-domain Modulation Pro- gressive Fusion Network (BFMPF-Net). Specifically, we propose a Bidirectional Frequency-domain Modulation Enhancement (BFME) module that effectively exploits bidirectional modu- lation between high- and low-frequency components and learns the spatial weights of high-frequency features to attenuate noise and preserve crack edge details, thereby improving the performance of crack segmentation. Furthermore, the Progressive Guidance Fusion module serves as another core component of our framework. It leverages the spatial prior provided by the original low-resolution image to guide feature refinement via stepwise optimization from coarse contours to fine edges, thereby ensuring the integrity of crack segmentation. Evaluation on three publicly available datasets—CrackTree260, CrackLS315, and Crack760—affirms the superior segmentation accuracy of the proposed BFMPF-Net compared to current mainstream methods.