AGSENet: A Robust Road Ponding Detection Method for Proactive Traffic Safety

TL;DR

AGSENet incorporates saliency detection techniques through the Channel Saliency Information Focus (CSIF) and Spatial Saliency Information Enhancement (SSIE) modules and outperforms existing methods, setting a new state-of-the-art in this field.

Abstract

Road ponding, a prevalent traffic hazard, poses a serious threat to road safety by causing vehicles to lose control and leading to accidents ranging from minor fender benders to severe collisions. Existing technologies struggle to accurately identify road ponding due to complex road textures and variable ponding coloration influenced by reflection characteristics. To address this challenge, we propose a novel approach called Self-Attention-based Global Saliency-Enhanced Network (AGSENet) for proactive road ponding detection and traffic safety improvement. AGSENet incorporates saliency detection techniques through the Channel Saliency Information Focus (CSIF) and Spatial Saliency Information Enhancement (SSIE) modules. The CSIF module, integrated into the encoder, employs self-attention to highlight similar features by fusing spatial and channel information. The SSIE module, embedded in the decoder, refines edge features and reduces noise by leveraging correlations across different feature levels. To ensure accurate and reliable evaluation, we corrected significant mislabeling and missing annotations in the Puddle-1000 dataset. Additionally, we constructed the Foggy-Puddle and Night-Puddle datasets for road ponding detection in low-light and foggy conditions, respectively. Experimental results demonstrate that AGSENet outperforms existing methods, achieving IoU improvements of 2.03\%, 0.62\%, and 1.06\% on the Puddle-1000, Foggy-Puddle, and Night-Puddle datasets, respectively, setting a new state-of-the-art in this field. Finally, we verified the algorithm's reliability on edge computing devices. This work provides a valuable reference for proactive warning research in road traffic safety.

Figures (15)

• Figure 1: Typical traffic accidents caused by road ponding, from top to bottom, they are rollovers, skidding, and brake failure 919293949596.
• Figure 2: The framework of the proposed Self-Attention-based Global Saliency Enhancement Network (AGSENet) for Ponding-Water detection. The architecture incorporates an encoder, a decoder, a Channel Saliency Information Focus (CSIF) module (c), and a Spatial Saliency Information Exploration (SSIE) module (d). The encoder and decoder are built by stacking RSU-X (b), forming a U-shaped structure. The system is predicated on an encoder-decoder architecture to refine the segmentation of ponding water, with the CSIF module deployed to focus on salient information along the feature channel dimension. Concurrently, the SSIE module is harnessed to explore the spatial dimension's significant information across both high-level and low-level features.
• Figure 3: The structural diagram of the Channel Saliency Information Focus (CSIF) module includes Spatial Context Information Perception and Channel Saliency Information Interaction. The symbols Q, K, and V respectively represent Query, Key, and Value, while H, W, and C denote the shape of the features.
• Figure 4: The responses of different channels to specific object categories are shown from left to right as trees, sky, water, and ground, respectively. The red boxes represents the reflection of objects in the water in a real-world scene.
• Figure 5: The structural diagram of the Spatial Saliency Information Exploration (SSIE) module includes ambiguous feature noise removal and edge position refinement. H, W, and C denote the shape of the features.