It is a challenging task to accurately segment damaged road markings from images, mainly due to their fragmented, dense, small-scale, and blurry nature. This study proposes a multi-scale spatial kernel selection net named M-SKSNet, a novel model that integrates a transformer and a multi-dilated large kernel convolutional neural network (MLKC) block to address these issues. Through integrating multiple scales of information, the model can extract high-quality and semantically rich features while generating damage-specific representations. This is achieved by leveraging both the local and global contexts, as well as self-attention mechanisms. The performance of M-SKSNet is evaluated both quantitatively and qualitatively, and the results show that M-SKSNet achieved the highest improvement in F1 by 3.77% and in IOU by 4.6%, when compared to existing models. Additionally, the effectiveness of M-SKSNet in accurately extracting damaged road markings from images in various complex scenarios (including city roads and highways) is demonstrated. Furthermore, M-SKSNet is found to outperform existing alternatives in terms of both robustness and accuracy.

To tackle the complexity and limited applicability of high-precision segmentation models for damaged road markings, this study proposes a Multi-level Adaptive Lightweight Network (MALNet) based on knowledge distillation. By incorporating multi-scale dilated convolution and adaptive spatial channel attention fusion modules, the MALNet model significantly enhances the precision, integrity, and robustness of its segmentation branch. Furthermore, it employs an intricate knowledge distillation strategy, channeling rich, layered insights from a teacher model to a student model, thus elevating the latter’s segmentation ability. Concurrently, it streamlines the student model by markedly reducing its parameter count and computational demands, culminating in a segmentation network that is both high-performing and pragmatic. Rigorous testing on three distinct data sets for damaged road marking detection—CDM_P (Collective Damaged road Marking—Public), CDM_H (Collective Damaged road Marking—Highways), and CDM_C (Collective Damaged road Marking—Cityroad)—underscores the MALNet model’s superior segmentation abilities across all damage types, outperforming competing models in accuracy and completeness. Notably, the MALNet model excels in parameter efficiency, computational economy, and throughput. After distillation, the student model’s parameters and computational load decrease to only 31.78% and 27.40% of the teacher model’s, respectively, while processing speeds increase to 1.9 times, demonstrating a significant improvement in lightweight design.